WO2005094480A2

WO2005094480A2 - Power assist device

Info

Publication number: WO2005094480A2
Application number: PCT/US2005/009593
Authority: WO
Inventors: Edwin V. Slate; Dana R. Yoerger; Richard J. Resch; Timothy E. Thiel; Nathan T. Ulrich
Original assignee: Motiv Technology, Inc
Priority date: 2004-03-23
Filing date: 2005-03-21
Publication date: 2005-10-13
Also published as: WO2005094480A3

Abstract

A power assist device (10) to assist a user in effecting movement of a wheeled device. The power assist device has a user interface (12) configured to be gasped by a user. The user interface measures up to six degrees of freedom forces and torque applied to the user interface and converts measured force and torque into electronic signals. The power assist device includes a controller (14) having a processing resource to process the electronic signals. The controller provides at least one velocity reference signal representing a desired velocity. The velocity servo has an input for receiving the velocity reference signal and outputs a velocity feedback signal representing an actual velocity. The processing resource is configured to process the electronic signals outputted by the user interface and the velocity feedback signal to generate the velocity signal.

Description

DESCRIPTION

POWER ASSIST DEVICE

Technical Field The present invention generally relates to a power assist device. Background Art Power assist devices and vehicles are known in the art and are described in U.S. Patent Nos. 6,459,962, 6,302,226, 6,230,831, 6,202,773, 5,818,189 and 4,339,013.

Disclosure of the Invention The present invention is directed to a power assist device that is applicable to such devices as lawn mowers, carts, hospital gurneys, etc. In accordance with the invention, the power assist device comprises a user interface and a power assist controller. The user interface measures the force and/or torque interaction between the user and the power assist device. The input force and/or torque applied by the user to the user interface is measured and converted to electronic signals that are inputted into the power assist controller. The power assist device further comprises at least one velocity servo that has a drive motor, a transmission and a motor driven wheel. The input force imparted by the user on the user interface directly acts on the power assist device, either in parallel with or through the drive motor and transmission, and works in conjunction with the motive power developed by the motor driven wheels. In accordance with the invention, the power assist controller operates on the user input force and/or torque using one or more control algorithms to produce velocity references for the velocity servo. The algorithms employ one or more mathematical models of "ideal" devices, characterized by a desired mass parameter and desired drag law and value. The mathematical model of the ideal device is used in conjunction with the measured user input forces and/or torques and the actual velocity to produce a desired velocity for the velocity servos and can also command an additive torque (feed-forward torque) . Standard servo loop technology, using an electric motor, or hydraulic or pneumatic actuators, is employed to minimize the error between desired velocity and actual velocity for each velocity servo. A power supply provides electrical power to the user interface, the power assist controller and each velocity servo. The power supply may be configured as a battery bank, or an engine driven alternator in conjunction with a voltage regulator and battery. Thus, in one embodiment, the present invention is directed to a power assist device to assist a user in effecting movement of a wheeled device, comprising a user interface configured to be grasped by a user and measure up to six degrees of freedom forces and torque applied to the user interface by the user and convert the measured forces and torque into electronic signals. The power assist device includes a controller having a processing resource to process the electronic signals. The controller provides at least one velocity reference signal representing a desired velocity. The power assist device further comprises at least one velocity servo having an input for receiving the at least one velocity reference signal wherein the at least one velocity servo outputs a velocity feedback signal representing an actual velocity. The processing resource of the controller is configured to process the electronic signals outputted by the user interface and the velocity feedback signal to provide the at least one velocity signal. In a related aspect, the present invention is directed to a method for assisting a user in effecting movement of a wheeled device comprising the steps of providing a wheeled device having a user interface configured to be grasped by a user and at least one velocity servo for assisting in the movement of the wheeled device wherein the user interface is configured to measure forces and torque applied thereto by a user, and measuring up to six degrees of freedom forces and torque applied to the user interface. The method further comprises the steps of converting the measured forces and torque into electronic signals, processing the electronic signals to provide at least one velocity reference signal representing a desired velocity, and inputting the at least one velocity reference signal into the at least one velocity servo. The method further comprises generating a velocity feedback signal representative of an actual velocity of the wheeled device, and processing the velocity reference signal and the velocity feedback signal to minimize error between these signals and provide an updated velocity reference signal.

Brief Description of the Drawings The features of the invention are believed to be novel. The figures are for illustration purposes only and are not drawn to scale. The invention itself, however, both as to organization and method of operation, may best be understood by reference to the detailed description which follows taken in conjunction with the accompanying drawings in which: FIG. 1 is a block diagram of the power assist device of the present invention; FIG. 2 is a diagram illustrating three-axis force measurement provided by a sensor of the power assist device of the FIG. 1; FIGS. 3A and 3B are diagrams illustrating forces that are applied to a handle that is part of a user interface of the power assist device of the present invention; FIG. 4A is a top plan view of a flexure device utilized by the power assist device of the present invention; FIG. 4B is a cross-sectional view along line 4B-4B of FIG. 4A; FIG. 4C is a diagram illustrating the movement of a flexure member shown in FIG. 4A in response to a force applied thereto; FIG. 5 is a block diagram of a power assist controller shown in FIG. 1; FIG. 6A is a diagram illustrating a two-wheel vehicle and a corresponding base coordinate frame; FIG. 6B is a side view taken along line 6B-6B of FIG. 6A; FIG. 7 is a diagram illustrating a corresponding sensor coordinate frame in conjunction with the two-wheel vehicle and base coordinate frame shown in FIG. 6A; FIG. 8 is a block diagram illustrating the algorithm implemented by a central processing unit shown in FIG. 5; FIG. 9 is a block diagram of a dynamic model shown

in FIG. 5; FIG. 10 is a diagram illustrating two-dimensional drive velocity allocation; FIG. 11 is a block diagram illustrating a power generation scheme to supply power to the power assist device in accordance with one embodiment of the present invention; FIG. 12A is a flow chart illustrating a power assist supervisory control methodology in accordance with another embodiment of the invention; FIG. 12B is a diagram illustrating hand truck pitch angles; FIGS. 13A and 13B are diagrams that illustrate the coordinate frame and force definitions used with respect to a two-wheeled device; FIG. 14 is a plot that shows three curves for the pitch moment as a function of pitch angle for identical load weights with different locations for the center of gravity (CG) of a two-wheeled device; FIG. 15 is a plot that shows experimental data for the moment, as measured at the instrumented handle of the two-wheeled device, as a function of pitch angle for three different loads; FIG. 16 is a diagram that shows the additional dimensions needed to specify the moment supplied by the user of the two-wheeled device; and FIG. 17 is a block diagram that illustrates implementation of the method of the present invention on a two-wheeled device wherein such implementation is based on the linear model case.

Modes For Carrying Out The Invention The power assist device of the present invention can be applied to many single and multi-wheel devices such as wheelbarrows, lawn mowers, hand trucks, carts, hospital gurneys, etc. Referring to FIG. 1, there is shown a block diagram of power assist device 10 in accordance with one embodiment of the present invention. Power assist device 10 generally comprises user interface 12, power assist controller 14 and velocity servos 16A-N. Thus, it is to be understood that although FIG. 1 shows up to N velocity servos 16, there can be just one velocity servo, or there can be a plurality N velocity servos. User interface 12 measures the force and/or torque interaction between the user and power assist device 10. As will be described in the ensuing description, user interface 12 is part of or integral with the handle or handles of a wheeled device with which power assist device 10 is used. The handle or handles are grasped by a user or operator. User interface 12 converts the measured input force and/or torque applied to the handle by the user into electronic signals. These electronic signals are inputted into power assist controller 14. Power assist controller 14 operates on the user input force and/or torque using one or more control algorithms to produce velocity reference signals 17A-N that are inputted into corresponding velocity servos 16A-N, respectively. The algorithms employ one or more mathematical models of "ideal" devices, characterized by a desired mass parameter and desired drag law and value. Such suitable algorithms are described in U.S. Patent No. 6,459,962, the disclosure of which patent is incorporated herein by reference. Each velocity servo 16A-N outputs a corresponding velocity feedback signal 19A-N, respectively. The mathematical model of the ideal device is used in conjunction with the measured user input forces and/or torques and the actual velocity (e.g. velocity feedback signals 19A-N) to produce a desired velocity (e.g. velocity reference signals 17A-N) for velocity servos 16A-N. The mathematical model of the ideal device is also used in the presence of an additive torque (feed-forward torque) . In one embodiment of the invention, each velocity servo 16A-N utilizes standard servo loop technology to minimize the error between desired velocity and actual velocity for each velocity servo 16A-N. In one embodiment of the invention, each velocity servo 16A-N comprises motor amplifier 20, electric drive motor 22, transmission 24, drive wheels 26 (i.e. motor driven wheels) or hydraulic or pneumatic actuators (not shown) . Thus, the input force imparted by the user on user interface 12 directly acts on the power assist device either in parallel with or through the electric drive motor 22 and transmission 24, and works in conjunction with the motive power developed by the drive wheels 26. Power supply 30 provides electrical power to user interface 12, power assist controller 14 and each velocity servo 16A-N. Power supply 30 can be configured as a battery bank or an engine driven alternator in conjunction with a voltage regulator and battery. Referring to FIG. 1, user interface 12 measures the forces and/or torque interaction between the user and the power assist device and converts the measured forces and/or torque to electronic signals that are inputted to power assist controller 14. The forces and/or torques measured by user interface 12 are processed by power assist controller 14 to generate the velocity reference signals 16A-N that are inputted into velocity servos 16A- N, respectively. In one embodiment, user interface 12 measures the six degrees of freedom (DOF) forces and torques applied by the user or operator. The user or operator grasps the handle (which is part of user interface 12), and the forces and/or torques applied to the handle are measured by a force/torque sensor of user interface 12. (The term "sensor handle" is used in the ensuing description to refer to a handle having the aforementioned sensor) . For devices in which pitch and roll can vary (i.e. devices without rigid frames or with fewer than three wheels on the ground) , the pitch and roll of the device must be measured in order to transform the measured forces and moments into a frame that is fixed relative to the gravity vector. In many cases, the full six DOF forces and torques applied at the sensor handle will not be required by power assist controller 14. In most power assist applications, measurement of the two or three components of orthogonal forces at the sensor handle will meet the requirements of the application. For devices with rigid or suspended frames and three or more wheels on^" the ground (such as a cart or hospital gurney) , the force in the forward/aft direction must be measured to control forward/aft motion and either the lateral force at the sensor, the moment about the vertical axis of the sensor, or both, must be measured to control steering. If the device has lateral mobility, then the lateral force at the sensor will be measured to control the lateral motion. For a fixed stability device of this nature, pitch and roll measurements are not needed since the sensor remains nearly fixed in orientation relative to gravity. For devices with two wheels, such as a hand truck, the net force in the forward/aft direction must be measured in the presence of variable orientation relative to gravity. Typically, this involves transforming two axes of force at the sensor handle into a horizontal force using a measure of the pitch of the hand truck. Steering control is enacted through either lateral force at the handle or moment about the vertical axis at the handle. In general, the device pitch measurement will be required to transform the handle sensor measurements. In the most general case, the sensor that is used to measure the forces and torques at user interface 12 is configured to measure the three orthogonal components of force at the handle of user interface 12. Other measurement configurations can be developed from the general case via reconfiguration of the basic elements and methodologies described herein. Referred to FIG. 2, there is shown a simplified representation of a sensor handle that provides the necessary measurement of three-axis forces. Handle 100 is attached to triangular handle plate 101. At each vertex of the triangular handle plate 101, a spring element 103 is coupled to the matching vertex of an opposing triangular base plate 102. A total of three spring elements 103 connect the handle plate 101 to the base plate 102. In practice, base plate 102 is attached to the power assist device. If spring elements 103 are constrained to provide motion only in the direction of positive and negative X axis as shown, then forces in the positive X direction applied to handle 100 will generally compress spring elements 103. Forces applied in the negative X axis direction will extend spring elements 103. Forces applied to handle 100 in the positive vertical direction (positive Z axis direction) will cause the upper spring elements to compress, and the lower spring element to extend due to the moment imparted by the vertical force. Forces applied to handle 100 in the negative vertical direction will cause the reverse, i.e., the upper spring elements will extend and the lower spring element will compress. Finally, lateral forces applied at handle 100 in the positive or negative Y axis direction will cause compression of the upper spring element in the direction of the lateral force and extension of the other upper spring element due to the moment imparted by the lateral force. Combinations of forces applied in the X, Y and Z axis directions will in general cause spring elements 103 to compress or extend depending on the magnitude and direction of the net applied force. If the displacement exhibited during compression or extension of the spring elements 103 at each vertex of handle plate 101 is measured at various levels of known applied force, then a calibration can be developed for each spring element 103. This calibration would provide a relationship between applied force and displacement for each spring element 103. Using the calibration for each spring element 103 when an unknown force is applied to handle 100, the measured displacements of the three spring elements 103 will yield the forces at each of the attachment points for spring elements 103. For a given geometry of handle 100 and handle plate 101, knowledge of forces at the three locations will allow computation of the three orthogonal force components. Referring to FIGS. 3A and 3B, handle 100 and handle plate 101 are illustrated in isometric view. The three known forces at the vertices of the handle plate 101 are labeled as FI, F2 and F3. FIG. 3B shows a side view of handle 100 and handle plate 101, and the forces FI, F2 , and F3 acting on handle plate 101. The external forces applied to handle 100 by the user are illustrated as Fx and Fz, representing the two orthogonal components of the external force to be measured. For the purposes of facilitating understanding of this aspect of the invention, it is assumed that the third component of force, Fy, is zero (this assumption in no way affects the validity or generality of the approach) . The handle geometry is identified as follows: dl is the height between the application points of the two top forces FI and F2, and the normal to handle plate 101 through the external force, d2 is the height between the bottom force application point and the normal to handle plate 101 through the external force, and w is the distance of the plate normal to external force. The external forces Fx and Fz are computed using statics as shown by equations ( 1 ) and ( 2 ) :

Fx = Fi + F₂ + F₃ ( 1 ) F_z = ( F₃ x d2 - ( Fi + F₂ ) x dl ) ÷ w ( 2 )

In the foregoing discussion, the assumption was made that spring elements 103 are constrained to motion only in the direction of the positive and negative X axis direction (see FIG. 2) . This requires that the compliance of the spring elements 103 be infinitely stiff in the Y axis and Z axis directions. Further, if handle plate 101 is rigid with fixed dimensions between spring element attachment points, the motion of handle plate 101 will be limited to pure translation motion with all spring attachment points moving the same distance. This is due to the fact that any differential in motion between the spring attachment points results in a lengthening or shortening of the distances between attachment points. Physically, this can occur only if the handle plate dimensions change between attachment points, or motion of the spring attachment points in the Y and Z axis directions is permitted. Additionally, differential motion between the spring attachment points will require that the angle between spring elements 103 and handle plate 101 be unconstrained, for example, as would be the case if the attachments to handle plate 101 were made using ball joints. One method to allow differential motion of the spring element attachment points to handle plate 101 in the presence of the infinitely stiff springs in the off- axis directions and the fixed dimensions of handle plate 101 is to provide for some nominal compliance at the attachment point. Using compliant elements, similar to engine isolation mounts, between spring elements 103 and handle plate 101 permit the necessary dimensional changes and rotational freedom. Referring to FIGS. 3A and 3B, spring elements 103 are required to exhibit desired compliance characteristics for the intended direction of motion yet provide near infinite compliance for off-axis motions. In accordance with the invention, a flexure device that exhibits these characteristics is shown in FIGS. 4A and 4B. The flexure device comprises housing 104, flexure member 105, force attachment 106 and flexure root attachment 107. Housing 104 defines recess 108 within which flexure member 105 flexes. Specifically, flexure member 105 flexes under the influence of externally applied force F. Flexure member 105 is a blade type element with the blade root attached to housing 104 via attachments 107. Flexure member 105 has end 105a. The force is directed to end 105a of flexure member 105 via force attachment 106. Recess 108 allows flexing of flexure member 105 without contacting housing 104. The thickness T of flexure member 105 is selected to provide the desired compliance characteristics. The height h of flexure member 105 is selected to provide the desired high off-axis stiffness (force stiffness and torsional stiffness). Housing 104 serves to provide a method for attaching the root of flexure member 105 as well as a means of providing overload protection. Referring to FIG. 40, the inner recess 108 in housing 104 is designed to limit the travel of flexure member 105 at some predetermined force, and therefore displacement value. This overload protection mechanism prevents plastic deformation of flexure member 105 or damage to the sensing elements under loads that exceed the nominal sensing range. Specifically, housing recess 108 is shaped to permit flexure member 105 to flex without contacting housing 104 for a range of deflections up to displacement do- At displacement do, occurring at a force of F₀, flexure member 105 contacts the housing walls defining housing recess 108 for a fraction of the length of flexure member 105. For applied forces greater than F₀, the effective flexure length is diminished, and the stiffness increases considerable. Thus, the shape of recess 108 can also be used to shape the resulting compliance characteristics of flexure member 105. A typical force-versus-displacement relationship for flexure member 105 is also shown in FIG. 40. At displacement d₀, the compliance (slope) changes due to the effective shortening of the flexure length. This characteristic is very desirable for the power assist application since the sensor can be made very sensitive at low forces (high sensitivity defined as large displacement per applied force) , yet provide a wide sensing range (at reduced sensitivity) for large forces. The displacement of flexure end 105a can be measured using commercially available non-contact displacement measurement technology, such as linear hall-effect devices, optical devices, or capacitive measurement devices. For example, in one embodiment, one or more magnets can be mounted to flexure end 105a, and a linear hall-effect integrated circuit can be mounted to housing 104 in proximity to the magnets. Such a configuration results in a near linear voltage-versus-displacement relationship. As a result of calibrating the transducer output voltage or current as a function of known values of applied weights, a calibration curve for each of the flexure based spring elements can be developed. Programming the calibrations into a programmable central processor unit (CPU) , which is described in the ensuing description, permits the sampled voltages or currents from each flexure sensor to be turned into a force magnitude value. The direction of the applied force for small motions of a typical flexure is normal to the direction of flexure motion. Thus, given a fixed mounting configuration for the flexure members, the direction of the applied force is fixed relative to the frame or body of the power assisted device. Referring to FIG. 5, there is shown a block diagram of power assist controller 14. Power assist controller 14 is similar to the power assist controller described in U.S. Patent No. 6,459,962, which patent was cited in the foregoing description. However, power assist controller 14 is configured to operate in three dimension scenarios, covering devices that are capable of six degrees of freedom (DOF) motion in 3-D space rather than planar motion alone. Power assist controller 14 generally comprises central processing unit 200 (CPU) , user interface force transducer 202, orientation transducer 203, motor velocity transducer (s) 204, and dynamic model 205. In one embodiment, central processing unit 200 further comprises external computer 209 that is connected to CPU 200 via wireless Ethernet link. The purpose of computer 209 is described in the ensuing description. Power supply 30 (see FIG. 1) supplies power to CPU 200 and associated control equipment. CPU 200 provides velocity reference signals 17A-N which are inputted into velocity servos 16A-16N, respectively, as described in the foregoing description. In a preferred embodiment, CPU 200 comprises a programmable microprocessor. CPU 200 provides the control system operation. Circuitry for the control system is preferably housed in an electronics control box (not shown) which is integrated with or mounted to the frame of power assist device 10. CPU 200 interfaces to user interface force transducers 202 and orientation sensor 203, and computes velocity reference signals 17A-N for velocity servos 16A-N in response to measured forces and/or torque at user interface 12. Dynamic model 205 provides dynamic representation of a "reference device" with desired mass and damping properties and is used to generate the velocity references from the user interface forces. The closed loop velocity control uses the error between the velocity reference provided by dynamic model 205 and the measured velocity of motor 22 (see FIG. 1) to control the output of motor 22 such that the error component is reduced to zero under steady state conditions. Referring to FIG. 5, in one embodiment, CPU 200 is interfaced to external computer 209 to provide a two-way wireless data path for reprogramming the software of CPU 200, as well as for monitoring the real time operating conditions of the control system. Preferably, external computer 209 comprises the software and hardware necessary for generating the executable control program that runs in CPU 200. In one embodiment, external computer 209 comprises an executable program that is used to provide a real time display of the operating conditions of the control system in graphical or tabular form, for monitoring and control of real time control system parameters such as gains, and for recording the operating conditions for data analysis and review at a later time. A single three dimensional (3-D) mathematical model is used to represent the dynamics of the desired reference device. The 3-D model represents the dynamic behavior of the device in a fixed coordinate frame attached to the device. Referring to FIGS. 6A and 6B, the coordinate frame, referred to as base coordinate frame 211, has an orientation that is fixed relative to an earth reference. A common reference frame used for this representation is a Cartesian coordinate frame with the XY plane normal to the local gravity vector and the Z axis positive in the down direction. Selection of the fixed location of the frame relative to the power assist device is not critical, although it is usually chosen to simplify the dynamic model representation. Figure 6A shows a device having two wheels 210. The base coordinates of base coordinate frame 211 are chosen with the Y axis through the center of rotation of wheels 210, and the X axis at the mid-point of the wheel-to-wheel spacing. All measured forces applied to the two-wheeled device through the user interface 12 are resolved into a set of equivalent forces and torques acting on the 3-D model in the base coordinate frame 211. The forces and torques will generate translational and rotational accelerations (respectively) on the 3-D model based on the model's mass and moment of inertia and the net forces acting on the 3-D model. The net forces are the vector sum of the forces applied via user interface 12 (see FIG. 1) and the parasitic drag forces resulting from the desired damping values contained in the model. As described in the foregoing description, the force transducers in user interface 12 are comprised of one or more flexure based sensors (e.g. the flexure device shown in FIGS, 4A and 4B) that generate a voltage or current signal as a function of the applied force. The relationship between force and associated output voltage or current is determined by bench or laboratory calibration using weight standards. A single flexure device will provide measure of a linear force at a known geometric relationship to the flexure structure. Thus, the flexure device shown in FIGS. 4A and 4B is preferably located on power assist device at a known and measurable location and orientation relative to some fixed reference (s) on the wheeled device. Thus, the flexure device output will provide measurement of a single force magnitude with known location and orientation relative to the fixed reference (s) on power assist device 10. Referring to FIG. 7, sensor coordinate frame 212 is used to represent the forces F, indicated by reference number 213, applied via user interface 12. Sensor coordinate frame 212 is a Cartesian coordinate frame that remains attached to the sensors, maintaining a fixed orientation relative to user interface 12. As illustrated in FIG. 7, the axes for the sensor coordinate frame 212 are labeled as X_s, Y_s, and Z_s. As the device orientation changes relative to the fixed orientation base coordinate frame 211, the orientation of the sensor coordinate frame 212 also changes relative to the base coordinate frame 211. This is illustrated in FIG. 7 for

two values of device pitch angle θi and θ₂. At the first

value of device pitch angle θi corresponding to the

vertical orientation, the orientation of the sensor coordinate frame 212 is identical to the orientation of the base coordinate frame 210. As the pitch angle

changes to the second pitch angle θ , the orientation of

the sensor coordinate frame 212 changes relative to the orientation of base coordinate frame 210. In order to represent the applied forces F, indicated by reference number 213, in base coordinate frame 211, the data must be transformed from sensor coordinate frame 212. A three dimensional coordinate transformation is used to translate and rotate the applied force F 213 from sensor coordinate frame 212 to base coordinate frame 211. In general, the transformation will be a function of the orientation provided by the orientation transducer 203. Referring to FIG. 5, orientation transducer 203 is an accelerometer-based instrument that responds to the local gravity vector, providing a pitch and roll angle relative to the vertical reference. If the accelerometer-based instrument is subject to other accelerations in addition to gravity, the accelerometers in the transducer can lead to dynamic errors in pitch and roll measurement. Compensation for such errors can be accomplished using turning rate sensors about the pitch and roll axes, then combining the accelerometer measurements and turning rate measurements using a Kalman filter. This technique is used by several commercially available orientation sensors. The transducer raw accelerations and turning rates are inputted to CPU 200 via analog-to-digital converters (see FIG. 8), then processed by CPU 200 to produce the pitch and roll angles relative to the vertical reference. Simplified versions of this configuration based on uncompensated acceleration measurements may be used in cases where the acceleration errors are small. Such simplified configurations may also be based on other static tilt measurement methods, such as electrolytic transducers, pendulum type transducers, and others. Referring to FIG. 8, there is shown a block diagram of the algorithm that resides in CPU 200. In the generalized case, the 3-D dynamic model 205 (see FIG. 5) operates on the six degree-of-freedom (6-DOF) forces and torques defined in base coordinate system 211, producing a 6-DOF velocity reference also defined in base coordinate system 211. Assuming rigid body dynamics, the velocity reference vector can be mapped to 3-D velocity vectors at the locations of the drive actuators. Depending on the actuator configuration (orientation relative to the rigid body) and capability (degrees of freedom) , each individual velocity reference signal 17A- 17N (see also FIG. 1) is developed from the 3-D velocity vector at each associated servo or drive location. In general, each velocity servo 16A-N may not be able to develop the full 3-D velocity reference due to limited degrees of freedom. For example, attempting to slide a power assisted hand truck laterally will generate a base coordinate Y axis component of velocity reference acting on the hand truck body. This will result in Y components of velocity reference at each of the actuator locations (wheel contact locations) . However, since the Y component of velocity cannot be delivered by the fixed orientation wheeled actuators used on the hand truck, this velocity component will be ignored (i.e. not responded to) . In other applications, using alternative actuator configurations (e.g., steered driven wheels), this velocity component may be used to provide lateral motive effort. Applications such as patient gurneys or pallet trucks are candidates for three degree of freedom power assist control that provides power assisted lateral motion in addition to power assisted fore/aft motion and turning motions. The ensuing description describes details of the processing algorithm in FIG. 8 for the generalized 3-D case. Orientation sensor 203 is sampled by analog-to- digital converter 250. An analog, anti-alias low pass filter (not shown) filters signals prior to being inputted into analog-to-digital converter 250 to prevent aliasing of the sampled signal when sampled at a fixed rate by analog-to-digital converter 250. The bandwidth of the aforementioned anti-alias low pass filter is based on the sample rate of analog-to-digital converter 250. Once converted to digital representations, processing resource 252, which is part of CPU 200, processes the sampled signals to provide transducer orientation Θ. Specifically, processing resource 252 converts the sampled signals to orientation in engineering units (degrees or radians) in the case of floating point math, or to scaled engineering units in the case of integer math. Similarly, for each force transducer 202 in user interface 12, the transducer voltage or current outputs are processed with a low pass, anti-alias filter and then sampled by analog to digital converter (s) 254. Processing resource 256, which is part of CPU 200, processes the sampled representations to provide forces

fs_t . Specifically, processing resource 256 converts the

resulting sampled representations to forces in engineering units (newtons) in the case of floating point math, or to scaled engineering units in the case of integer math. The resulting force variables are representations of the forces applied at user interface 12 and represented in sensor coordinate frames 212. In general, there will be one sensor coordinate frame 212

per force transducer 202. Forces fs_t and the calculated

orientation θ are inputted into processing resource 258

which is part of CPU 200. Processing resource 258 then

transforms each force fs_f from a sensor coordinate frame representation to the single base coordinate trame representation 211 (see FIG. 7) . In the transformation process, a force in sensor coordinates may, in general, result in forces and torques in the base coordinate frame, thus the force representation in base coordinates is a 6-DOF vector of forces and torques. After all force transducer values are sampled and transformed to base coordinates, processing resource 260 sums the 6-DOF

vectors to yield the effective net forces and torques f_mt

acting on the device (i.e. cart, hand truck, gurney, etc.) . Dynamic model 205 operates on the 6 DOF

force/torque vector f_ικl defined relative to the base

coordinate system, to produce a 6 DOF velocity reference

vector V_κf also defined in the base coordinate system.

This 6 DOF velocity reference vector V_ref is then inputted

into drive velocity allocation processor 261. Referring

to FIG. 9, the dynamic model 205 incorporates mass M and

damping B properties for the desired behavior of power assisted device 10. The 1/s term represents numerical integration that operates on the acceleration vector to

produce the velocity reference vector V_n/ . The

aforementioned M and damping B properties are defined for each of the six degrees of freedom for the 3-D device model. Referring to FIG. 9, the block diagram shown therein represents six equations wherein each equation corresponds to one of the six degrees of freedom of the 3-D Dynamic Model. The quantities represented in FIG. 9 are represented by equations 3-6:

J net J net ,x nel.y J m m net ,x m nel,y m„ ( 3 :

wherein equation (3) represents the 6 DOF vector of forces and torques in the base coordinate frame;

V r.ef = V ref ,x "ref,y ref,, °ref,_X ∞ ref ,y ^ω 'ref, z ) (

wherein equation (4) represents the velocity reference vector in the base coordinate frame, with translational

velocities, v, and rotational velocities, ω.

wherein equation (5) is a 6x6 diagonal matrix containing the inverse of the PA model mass and inertia.

wherein equation (6) is a 6x6 diagonal matrix containing the translational damping terms, B_t, and rotational damping terms, B_r. The indicated form of the inverse "mass" matrix shows identical mass terms (1/M) for the translational degrees of freedom, and identical inertia terms (1/J) for the rotational degrees of freedom. In a real physical system, the mass terms would be identical. However, in practice, maintaining independent terms for the masses permits the power assist model to be "adjusted" on a per axis basis. For example, in an application with full x-y motion capabilities (e.g., a floor polishing machine), the x component of mass could be made larger than the y component of mass to "suppress" the x axis (forward/aft) motions relative to the y axis (side to side) motions. The drive velocity allocation algorithm converts the

six DOF velocity reference vector V_nf defined in the base

coordinate system to velocity references (e.g. velocity reference signals 17A) for each of the actuators (i.e. servo) defined in the system. This is accomplished by assuming rigid body dynamics with known geometry of the velocity servos or actuators relative to the base coordinate system. For rigid body dynamics, the velocity at any point on the body can be described by the sum of the velocity of a reference point on the body, in this case the origin of the base coordinate system, and the velocity due to rotation of the body about the same reference point. In the general case, the velocity reference is the sum of the translation velocity plus the rotational velocity caused by the off- axis rotations. An example of the dynamic model equations and velocity allocation for a two-wheel device are described below. In this example, it is assumed that each of the two velocity servos are capable of delivering velocities in the base coordinate X axis direction only. Under this assumption, the velocity reference vector X translational

component (v_reffX) and Z axis rotational component (ω_ref,z)

are the only elements of the velocity reference vector of which can generate Velocity Servo X axis motions. Due to the diagonal characteristic of the mass and damping matrices of the dynamic model, the force/torque vector X

force component (f_net,x) and Z axis moment (m_net,_z) can

generate the velocity reference vector X translational

component (v_ref/X) and Z axis rotational component (ω_ref,_z) • In this case , the Dynamic Model reduces to equations ( 7 ; and ( 8 ) :

Referring to FIG. 10, there is shown a diagram illustrating drive velocity allocation. The drive velocity allocation reduces to the two dimensional case shown. FIG. 10 shows a view of the base coordinate XY plane (in the direction of the positive Z base coordinate

axis) , and the X axis translation velocity V_{ref x} and the Z

axis rotational velocity ω_{nf z} . The individual wheel

velocity references are calculated from the X axis

translational velocity V_{ref x} and Z axis rotational

velocity_y_re/-_ and are represented by the set of equations ( 9 ) :

ax ,1 = V r. ef ,x + r ax ω ref ,z ( 9 ) = v ref ,x - r ax ,r 0) ref , ,∑

In the event any velocity reference exceeds the maximum value, it is limited to the maximum value. In this case, all other non-saturated velocity references are limited in magnitude to preserve the direction of the original net force vector in base coordinates. As described in the foregoing description and as shown in FIGS. 1 and 5, power assist device 10 contains one or more velocity servos 16A-N, each comprising motor amplifier 20, electric drive motor 22, transmission 24 and drive wheel 26. Each velocity servo 16A-N receives a velocity reference signal 17A-N, respectively, from power assist controller 14, and operates on the velocity reference signals 17A-N so as to minimize the error between actual motor velocity and velocity reference. The actual motor velocities produced by velocity servos 16A-N are represented by velocity feedback signals 19A-N, respectively. In a typical embodiment, motor amplifier 20 employs standard servo control techniques to provide closed loop velocity control of the motor velocity. The motor velocity servo loop regulates the motor torque to maintain the desired velocity reference in the presence of external disturbances to the velocity, such as drag, friction and changing gravity loads due to sloped terrain. As described in the aforementioned U.S. Patent No. 6,459,962, each motor amplifier 20 operates on a corresponding velocity reference signal (e.g. velocity reference signal 17A) and actual motor velocity (e.g. velocity feedback signal 19A) using a proportional plus integral (PI) compensator to develop the current loop (i.e., torque) reference. Motor amplifier 20 utilizes a closed loop current control inside the closed loop velocity control to maintain the current (torque) at the desired value. The current loop employs a proportional only or proportional plus integral control operating on the current error (difference between the current reference and actual current) to develop the motor voltage command. The motor voltage command signal is a pulse width modulated duty cycle, which is then used to control the power amplifier stage of motor amplifier 20. In one embodiment, this power amplifier stage comprises a PWM H-bridge amplifier. The measured motor velocity may be obtained with a velocity sensor, either an optical encoder or conventional tachometer mounted to the motor shaft. In the case of brushless motors, the sensors used for motor commutation (typically optical or magnetic hall-effect) may be used for providing the measured motor velocity. Optionally, sensor-less measurement may be employed, in which case the motor velocity is calculated using an estimate of the motor back-EMF (BEMF) . This estimate is based on measurements of motor current and voltage, and a mathematical model of the motor's electrical and magnetic parameters. Drive motor 22 can be any actuator suitable for providing the required motive power for the device. Typically, brush or brushless DC motors, hydraulic motors, or pneumatic motors may be used depending upon the specific requirements of the application. The details of motor amplifier 20 are similar in each case, although the power amplifier stage of motor amplifier 20 may be different depending on the motor type used. For example, a brush motor will employ a conventional four- transistor H bridge, with commutation provided by the motor, whereas a brushless motor will employ one H bridge per motor phase, and the commutation is provided by motor amplifier 20 in conjunction with motor mounted commutation sensors. The motor is directly connected to transmission 24 which is typically a reduction gearbox. The output of transmission 24 is connected to and drives drive wheel 26. The specific configuration of transmission 24 depends upon on the specific torque and speed requirements of the device. Standard reduction gear designs may include worm gears, helical gear sets or planetary gear sets. The transmission may also incorporate a belt and pulley reduction stage, either alone or in conjunction with gear reduction. Alternative transmission configurations may also be used depending on the specific requirements of the application. An example would be a continuously variable transmission (CVT) of which many different configurations exist as known art. The combination of a low friction motor and a low reduction ratio transmission provides minimal drive-train drag in the cases where the motor is un-powered, such as during periods where the battery power is low, or when power assist device 10 is turned off, or the device is free-wheeled without the power assist function enabled. Referring to FIG. 1, each velocity servo 16 can be designed to provide motive effort in a fixed (single) direction using conventional rotational servos coupled to wheels (or tracks) , or in continually variable directions using a directional (steered) driven wheel. The directional servo is applicable for devices in which 3- DOF operation is desired, permitting independent lateral motions in addition to the forward-reverse, and turning motions. U.S. Patent No. 6,474,434, entitled "Drive Wheel", the disclosure of which is incorporated herein by reference, describes a device that provides the described rotating as well as turning motions for a single velocity servo. In using a directional velocity servo such as this, the velocity reference from dynamic model 205 (see FIG. 5) will consist of a velocity reference vector defined in base coordinates, with both magnitude and direction. The magnitude component will dictate the rotational speed of the wheel whereas the direction component will dictate the steering direction for the velocity servo. Referring to FIGS. 1 and 5, power supply 30 supplies the necessary electrical power to power assist controller 14, user interface 12, and velocity servos 16. Electrical power is supplied from a power source such as a battery bank, or an engine driven alternator in conjunction with a voltage regulator and battery. If only a battery bank is used as the electrical power source, then it is preferable that a rechargeable battery with high discharge capacity at high current draw be used. Typical battery technologies that meet these requirements include wet cell lead acid or Nickel Metal Hydride batteries. The battery capacity can be chosen to match the desired operating time of the device assuming some nominal operating duty cycle. When the battery capacity is discharged to the point the device is no longer operational, the battery can be swapped for a fully charged battery, or charged in place. In devices wherein a fueled prime mover (fueled engine) is included in the system, for example in a lawn mower, the prime mover may be used to provide the electrical power. Referring to FIG. 11, the fueled prime mover 400 provides mechanical power to an alternator or generator 401 via a rotating shaft 400A. The alternator or generator 401 can be integrated with the prime mover or mounted to or in close proximity to the prime mover 400. Typical small horsepower fueled engines incorporate magnets that are embedded in the prime mover's flywheel, rotating in close proximity to alternator coils attached to the housing around the flywheel. For larger horsepower prime movers, the alternator or generator is usually mounted to the engine, and driven from the engine rotating shaft via pulley and belt arrangement. Alternator or generator 401 provides alternating or DC voltage to voltage regulator 402. The voltage regulator 402 may incorporate a solid state rectifier (diode rectifier) to convert the alternating voltage output of the alternator to a DC voltage. Voltage regulator 402 provides regulated output voltage to battery 403 in the presence of a changing electrical load on battery 403. Standard regulation techniques such as shunt or series regulation are employed to provide the voltage regulation function. The battery voltage is controlled to maintain safe operating voltages on battery 403, as well as to provide optimum voltage levels for battery charging. Referring again to FIG. 11, battery 403 provides an electrical buffer between the load and the prime mover driven alternator or generator 401. If alternator or generator 401 is sized to meet or slightly exceed the average electrical load conditions of power assist device 10, then it should be possible to maintain the battery in a fully charged state. During periods when the electrical demand of the load exceeds the power level of the alternator or generator output, the battery charge is diminished. During periods when the electrical demand of the load is less than the power level of the alternator or generator output, the battery charge is replenished. Power Assist Supervisory Control For many classes of power assisted devices, the power assist algorithm can be enabled at system power up, and remain enabled until shut down by the operator, or by device specific automatic safety shutdowns (e.g., hardware faults) . Regulatory requirements or prudent design for safe operation may dictate that certain classes of devices incorporate some form of "dead-man" switch to activate and deactivate the power assist mechanism. There are classes of power assist devices that do not fit into either of these two categories, and that require some form of "supervisory" control for the automatic activation and deactivation of the power assist mechanism. These classes of power assist devices include devices that exhibit intrinsically unstable operation around one or more points of operation. A typical example is a loaded, two-wheeled hand truck, which exhibits unstable operation at all pitch angles other than the vertical angle and the balance point angle. In accordance with another embodiment of the invention, the control algorithm implemented by power assist controller 14 (see FIG. 8) is modified to assist in the safe initiation and termination of the power assist operation of intrinsically unstable devices such as those described in the foregoing description. In the case of a two- wheeled hand truck, one scenario is to prevent power assisted operation until the operator has rotated the hand truck to a comfortable operating angle, and to terminate power assisted operation when the pitch angle of the hand truck gets too close to vertical. The following describes one specific supervisory control methodology that addresses this class of devices. In order to facilitate understanding of this methodology, the algorithm implemented by power assist controller 14 is configured for use with the hand truck model. However, it is to be understood that the methodology is applicable and extendable to other configurations exhibiting different stability characteristics. Referring to FIG. 12A, there is shown a flowchart of the supervisory control methodology for the automatic activation and deactivation of power assisted operation. The first step 501 is the power-up step with the hand truck being in a vertical position. In steps 502 and 503, the supervisory control will establish zero velocity reference for each wheel velocity servo, and enable each wheel velocity servo. Additionally, the power assist algorithm will not respond to force inputs from the handle sensor. The net effect is to maintain zero wheel speed via the velocity servos. As the operator pulls back on the handle of a loaded hand truck, the velocity servos will maintain zero wheel speed and prevent any rolling motion of the hand truck. Thus, the applied force to the handle will result in a pitch moment about an axis through the wheel axles resulting in the rotation of the hand truck and load. As the hand truck pitch angle changes, step 504 continually monitors the pitch angle, and when the angle exceeds some fixed value p₀, step 505 implements the dynamic model 205 (see FIG. 5) which will generate the velocity references from the applied forces at user interface 12. Step 506 determines when the pitch angle exceeds a predetermined fixed value pi. When the pitch angle is less than pi, the power assist dynamic model is deactivated and step 503 ramps the velocity references to zero. As shown in FIG. 12B, angle pi will be very close to the vertical so as to allow the power assist to remain active for a wide range of pitch angles, yet provide a controlled shutdown of the assisted operation when the user returns the hand truck to the vertical position for unloading or loading. The supervisory control methodology described will further benefit the user by allowing them to rotate the hand truck to and from the vertical position without the need to place the foot on the wheel axle, as is customary practice with non-power assisted hand trucks, in order to retain the bottom of the unit from rotating. Thus, in accordance with a further embodiment, the present invention is applied to two-wheeled devices such as two-wheeled carts and hand-trucks. A two-wheeled vehicle such as a hand-truck in a static (motionless) state is unstable (i.e. it will fall over) except for three conditions: a) the vehicle is lying flat; or b) the vehicle is completely upright and resting on its nose plate; or c) the vehicle it tilted at a specific angle called the "balance point", an angle of tilt where the net gravitational forces create zero net moment about the hand-truck axle. When the hand-truck is tilted at angles closer to the operator than the balance point, the hand-truck will tend to fall toward the operator thereby approaching condition (a) above. When the hand-truck is tilted beyond the balance point, the hand-truck will move toward the upright position, thereby approaching condition (b) . The balance point, while stable, represents a point with negative static stability. Any perturbation from the balance point will result in movement away from the balance point. These problems result from the fact that the operator must apply forces to the handle of the two- wheeled vehicle in order to balance the vehicle. These forces are in addition to the forces the operator must apply in order to make the vehicle move or change speed, including starting, maintaining or changing speeds, and slowing to a stop. In accordance with the invention, the following method is used to solve the aforementioned problems related to two-wheeled vehicles. For purposes of illustrating this embodiment, the two-wheel vehicle is a hand-truck. First, upon start-up, the operator must tilt the hand-truck over a sufficient angular range so that parameters of the combination of the hand-truck and load can be automatically determined. During operation, the forces applied by the operator are transformed into an appropriate coordinate frame and the tilt angle of the hand-truck is measured continuously. In the next step, the net forces required at the current angle of tilt to balance the hand-truck and load are computed using the model of the hand-truck and the load. Next, the forces required for balance are subtracted from the total forces applied by the operator to produce an estimate of the forces intended to propel the hand- truck. This difference is then used to drive the power assist algorithm described in the foregoing description (see FIG. 8) . As a result of this method, the operator can remain stationary with the hand-truck at any tilt angle and the hand-truck will remain stationary even though a non-zero set of forces is applied to the instrumented handle. These forces can be substantial depending on the load. This method does not use the control system to automatically balance the system. The method actually mimics a manual system, but with lower operator effort. Thus, although the method would require the operator to balance the system as for a manual hand- truck, the operator input forces required to propel or stop the hand-truck are significantly reduced. Thus, this method identifies a model of the combined hand-truck and load, and then uses the model in real time to remove the operator-input forces to balance the hand-truck from the force estimates that are sent to the power assist algorithm In order to implement this embodiment of the invention, a processing system, drive system and power supply module are attached to the two-wheeled vehicle. The processing system pre-processes the data signals defining the forces needed to balance and drive the two- wheeled vehicle and then implements the power assist algorithm and method, as described in the foregoing description and shown in FIG. 8, using the pre-processed

data. Each step of the method relating to two-wheeled vehicles described above is now described in detail. a) Model Identification Prior to implementing the power assist algorithm described in the foregoing description with respect to a hand-truck, the hand-truck operator must move the hand- truck through a range of tilt angles so that the static properties of the hand-truck and load can be determined. This model can then be used to remove the forces required to balance the hand-truck from the forces applied by the operator for vehicle movement. FIGS. 13A and 13B show the coordinate frame and force definitions used in the calculations and processing. The coordinate frame is centered on a point on the wheel below the axle and the X axis points outward and the Z axis points down parallel to the frame. Critical quantities are: P, the angle of pitch from the vertical; Θ, the complement of P; 0=90° corresponds to straight up; Li, the Z coordinate of the CG; L₂, the height of the hand truck; L₃, the X coordinate of the frame; L₄, the distance along the X axis from the frame to the center of gravity (CG) . The total moment, which varies with the pitch angle, is described by equation 10: M = WL_l ∞sθ-W(L₃ +L_t)sinθ (10)

wherein "W" is the weight, and Li, L₃, and L₄ describe the offsets of the center-of-mass of the combined hand-truck and load. Startup begins with the hand-truck at the stable, upright position. If this condition is not met, the power assist algorithm and method will not be executed. The operator must then tilt the hand-truck toward himself well past the balance point (typically approximately 60°). During this motion, user interface 12 records the pitch angle and the handle sensor forces. Referring to FIGS. 13A and 13B, the coordinate system is placed at the bottom of the wheel below the axle. The position of the center of gravity, CG, is described by Li, L₃, and L₄. The angle θ describes the angle of tilt and has a value of 90° when the hand-truck is upright and is zero when the hand-truck frame is parallel to the floor. These measurements are then used to create the model for the torque required to balance the hand-truck. In the simplest implementation, we approximate equation (10) above by equation (11) :

M = aθ + β (11) Given measurements of M and θ over a sufficient range (over approximately 60°, including the balance point), the coefficients α and β can be identified by linear least squares. The resulting linear relationship between pitch angle and moment represents an approximation to the actual behavior. We can also identify the groups of terms in equation (10) by linear least squares to provide a more precise model of the moment as a function of θ. We can rewrite equation (10) as equation (12):

While this equation is nonlinear, it is linear in the parameters δ and ε, which can be computed by linear least squares. The additional computational burden over fitting the linear model is very minor. Equation (12) also shows an important aspect of this method. Specifically, only the grouped terms δ and ε are identified, not the actual lengths and weight. Thus, the weight of the load and the location of its center of gravity are not known. However, at this step in the method, it is known that the center of gravity of the load lies somewhere along a line passing through both the balance point and the origin. The angle of this line is defined by equation (13) :

However, the weight of the load could be large with its center of gravity at a small distance from the origin, or it could be a relatively lighter weight with its center of gravity farther away from the origin. Statically, these are the same if the product of the distance from the origin and the weight is equal. However, dynamically they are different. Different loads will result in moment-pitch angle relationships with different balance points and different shaped curves. FIG. 14 shows three such curves for identical load weights with different locations for the center of gravity (CG) . The weight of the load is 50 kg. In each case, the center of gravity (CG) is 0.4 meters forward from the origin (L₃+L₄) but the vertical distances (Li) of the center of gravity (CG) are 0.0 meters (low), 0.5 meters (medium) and 1.0 meters (high) . The curves shown in FIG. 14 are not linear and intersect 0.0 moment (the balance point) at greatly different angles. These functions resemble straight lines about the balance point, indicated by (*) . Accordingly, the moment required to balance the hand- truck at a reasonable angle (for example 45 degrees) varies in both magnitude and sign. Referring to FIG. 15, there is shown a plot that shows experimental data for the moment, as measured at the instrumented handle as a function of pitch angle for three different loads. Each load has different weight and center of gravity, with the balance point (moment=0) ranging from about 10° to about 90°. The linear and nonlinear model fits are also shown in FIG. 15. The start-up of the test is shown in the measured moment curve on the left side. This start-up response is ignored when the model parameters are computed. The nonlinear model fits the data better and most importantly, it predicts the balance point better in one of the three conditions (a) , (b) and (c) discussed in the foregoing description. b) Use of the Model in Real Time In accordance with this embodiment of the invention, the moment-angle relationship is used in real-time to remove the component of the operator' s input force that balances the two-wheeled vehicle. FIG. 16 shows the additional dimensions needed to specify the moment supplied by the user, and FIG. 17 shows implementation of the power assist method and algorithm for the linear model case. FIG. 17 is a processing diagram that illustrates how the moment required to balance the load (Mi_oad) is computed and then removed from the overall measured moment inputted by the operator (M_0per) • As diagrammed in FIG. 17, the operator moment (M_0per) is calculated from the X and Z components of the measured sensor forces (F_Sχ _and Fsz respectively) in the sensor coordinate frame, and the associated moment arms X₀ and Z₀ from the handtruck origin (X, Z) . The load moment (M_Load) illustrated in FIG. 17 uses the linear model derived from the Model Identification phase described in section a)

above, specifically equation 11. The model parameters, α

and β are derived during the last "power assist" turn-on

cycle, and are used to generate the estimated static balance moment (M_Load) for the load as a function of pitch

angle θ. The net operator propulsion force is then

calculated as the difference between the balance moment for the load (M_Load) and the measured operator generated moment (M_oper) • Without loss of generality, the load moment could be computed using the non-linear model derived from the Model Identification phase described in section a) above, specifically equation 12. The principles, preferred embodiments and modes of operation of the present invention have been described in the foregoing specification. The invention which is intended to be protected herein should not, however, be construed as limited to the particular forms disclosed, as these are to be regarded as illustrative and exemplary rather than restrictive. Variations and changes may be made by those skilled in the art without departing from the spirit of the invention. Accordingly, the foregoing detailed description should be considered exemplary in nature and not as limiting the scope and spirit of the invention as set forth in the attached claims. What is claimed is:

Claims

1. A power assist device to assist a user in effecting movement of a wheeled device, comprising:

a user interface configured to be grasped by a user and measure up to six degrees of freedom forces and torque applied to said user interface by the user and convert said measured force and torque into electronic signals;

a controller having a processing resource to process said electronic signals, said controller providing at least one velocity reference signal representing a desired velocity;

at least one velocity servo having an input for receiving said at least one velocity reference signal, said at least one velocity servo outputting a velocity feedback signal representing an actual velocity; and

said processing resource of said controller being configured to process said electronic signals outputted by said user interface and said velocity feedback signal to provide said at least one velocity signal.

2. The power assist device according to claim 1 wherein the power assist device includes comprises a wheeled device.

3. The power assist device according to claim 1 further including a power supply to power said user interface, said controller and said at least one velocity servo.

4. The power assist device according to claim 1 wherein said at least one velocity servo comprises means to minimize error between said velocity reference signal and said velocity feedback signal.

5. The power assist device according to claim 1 wherein said user interface comprises a sensor handle.

6. The power assist device according to claim 5 wherein said sensor handle is configured to measure three orthogonal forces at said sensor handle.

7. The power assist device according to claim 6 wherein said sensor handle comprises flexure device means to generate electrical signals representing applied force and torque.

8. The power assist device according to claim 2 wherein said processing resource of said controller is configured to implement a three-dimensional, six degree-of-freedom mathematical model to represent the desired dynamics of the wheeled device.

9. The power assist device according to claim 1 wherein said user interface is configured to measure predetermined, initial, static balance parameters of a two-wheeled vehicle.

10. The power assist device according to claim 9 wherein said processing resource of said controller is configured to process said measured predetermined, initial, static balance parameters to determine the forces or torques required to balance said two-wheeled vehicle.

11. The power assist device according to claim 10 wherein said processing resource subtracts said forces or torques required to balance said two- wheeled vehicle from the total forces or torques applied to the user interface so as to provide estimates of the user-supplied propulsion forces and to prevent user- supplied balancing forces from causing motion of the system.

12. A method for assisting a user in effecting movement of a wheeled device comprising the steps of:

providing a wheeled device having a user interface configured to be grasped by a user and at least one velocity servo for assisting in the movement of said wheeled device, said user interface configured to measure forces and torque applied thereto by a user;

measuring up to six degrees of freedom forces and torque applied to said user interface;

converting said measured force and torque into electronic signals;

processing said electronic signals to provide at least one velocity reference signal representing a desired velocity; inputting said at least one velocity reference signal into said at least one velocity servo;

generating a velocity feedback signal representative of an actual velocity of the wheeled device; and

processing said velocity reference signal and said velocity feedback signal to minimize error between these signals and provide an updated velocity reference signal.

13. The method according to claim 12 wherein said step of measuring comprises generating electrical signals representing applied force and torque.

14. The method according to claim 12 wherein said step of processing said electronic signals to provide at least one velocity reference signal representing a desired velocity comprises processing said electronic signals with a three-dimensional mathematical, six degree-of-freedom model to represent the desired dynamics of the wheeled device.

15. The method according to claim 12 wherein said step of measuring includes measuring predetermined, initial static balance parameters of a two-wheeled vehicle.

16. The method according to claim 15 wherein the step of processing said electronic signals comprises processing said measured predetermined, initial static balance parameters to determine the forces required to balance said two-wheeled vehicle.

17. The method according to claim 16 wherein said step of processing said electronic signals further comprises subtracting said forces required to balance said two-wheeled vehicle from the total forces applied to the user interface so as to provide estimates of the user supplied propulsion forces to propel the two-wheeled vehicle and prevent user-supplied balancing forces from causing motion of the two-wheeled vehicle.