A non-holonomic, highly human-in-the-loop compatible, assistive mobile robotic platform guidance navigation and control strategy

The provision of assistive mobile robotics for empowering and providing independence to the infirm, disabled and elderly in society has been the subject of much research. The issue of providing navigation and control assistance to users, enabling them to drive their powered wheelchairs effectively, can be complex and wide-ranging; some users fatigue quickly and can find that they are unable to operate the controls safely, others may have brain injury re-sulting in periodic hand tremors, quadriplegics may use a straw-like switch in their mouth to provide a digital control signal.

Advances in autonomous robotics have led to the development of smart wheelchair systems which have attempted to address these issues; however the autonomous approach has, ac-cording to research, not been successful; users reporting that they want to be active drivers and not passengers. Recent methodologies have been to use collaborative or shared control which aims to predict or anticipate the need for the system to take over control when some pre-decided threshold has been met, yet these approaches still take away control from the us-er. This removal of human supervision and control by an autonomous system makes the re-sponsibility for accidents seriously problematic.

This thesis introduces a new human-in-the-loop control structure with real-time assistive lev-els. One of these levels offers improved dynamic modelling and three of these levels offer unique and novel real-time solutions for: collision avoidance, localisation and waypoint iden-tification, and assistive trajectory generation. This architecture and these assistive functions always allow the user to remain fully in control of any motion of the powered wheelchair, shown in a series of experiments.