Adaptive Motion Estimation and Control of Intelligent Walkers

摘要

One of the most critical factors in the quality of elderly lives is their ability to move. As
\nthe size of the ageing society grows, more elderly people suffer from walking impairments.
\nMost of them prefer to stay at home due to the shortage of the nursing care staff, since
\nthey deal with the daily challenges alone. Robotics researchers have developed various
\nintelligent walking support systems to meet the needs of elderly and handicapped people.
\nA particular problem in path tracking for such systems is maintaining the tracking performance,
\nwhich is affected by the center of gravity (CG) shifts and load changes due to
\nhuman-walker interactions. This thesis focuses on design of feedback controllers for safe
\nmotion of intelligent walker (i-walker) systems robust to CG shifts and load changes. Our
\ndesign follows a two level approach, one for kinematics, the other for dynamics. The high
\nlevel kinematic controller is designed based on integrator backstepping to produce desired
\nvelocities required for trajectory tracking. The low level dynamic controller is composed of
\na feedback linearization unit and a linear feedback controller to apply the control torque
\nfor tracking the desired velocity produced by the high level kinematic controller. As dynamic
\ncontrollers, proportional-derivative (PD) and sliding mode controllers (SMCs) are
\ndesigned. In our initial design, we assume that all system states are available. However, in
\nthe actual case, even if the wheel velocities can be measured with some sensor devices like
\ntachometers, the measurements carry noise, which poses important problems in control algorithms.
\nTo obtain the estimates of the wheel velocities, avoiding the noise problems, the
\ndesign of sliding mode observers and high gain observers is studied. The state feedback
\nPD and SMC schemes are later integrated with these observers to form implementable
\noutput feedback controllers. In practice, the human mass and the distance due to the CG
\nshift depend on the user. To address this issue, the output feedback control designs are
\nfurther made adaptive, integrating with a parameter identifier to estimate these variables.
\nThe parameter identifier design involves a linear parametric model of the i-walker system
\ndynamics and a least-square adaptive law based on this parametric model. Adaptive versions
\nof the above observers and control designs are done utilizing estimated parameters
\nand states. The effectiveness and applicability of the proposed controllers are verified via
\nvarious simulations in MATLAB/Simulink environment.