The challenge

In a society that does not provide unfettered wheelchair access in every public location, patients who are wheelchair-bound face barriers to normal daily movement that include the navigation of steps and curbs. This forces them to rely on the kindness of strangers to assist them to move around in public spaces. Well-meaning but inexperienced persons offering to help wheelchair-bound patients navigate steps, curbs and ramps only increase the risk of falls and injury.

To date, even powered wheelchairs have not been able to provide a solution. They rely on the user having the cognitive and fine motor skills required to control the chairs, and research indicates that around 10-40% of patients who could benefit from powered wheelchairs are refused one because they don’t have the skills to use it.

The proposed solution

This project aims to produce wheelchairs that can:

• predict human intention,
• navigate in complex environments, and
• travel on uneven terrain.

We harness AI technology to create a wheelchair planning and control system that uses a three-layer architecture:

• a high-level module for goal inference – uses probability-based Al algorithms to reason human intention under uncertainties, which can tolerate imprecise joystick control.
• a mid-level module for selection of trajectories – uses AI-based perception algorithms to allow the robot to recognise complex environments (such as door ways, walls of narrow corridors, steps, cubes, etc.) and then choose the optimal trajectory.
• a low-level module for obstacle avoidance – additional modular powered seat and crawl track mechanism is being developed to allow the wheelchair to auto-adjust balance on uneven floors and to climb curbs and stairs.