Digital sensor system for powered wheelchairs will use AI for driver assistance

Intelligent powered wheelchairs that adapt the level of support they offer users depending how skilled they are as a driver or how tired they become, could offer disabled people the opportunity for greater independence.

powered wheelchairs

Researchers at Portsmouth University, led by Dr David Sanders and Dr Martin Langner, have previously developed low-cost analogue collision avoidance systems for powered wheelchairs, based on the use of simple proximity sensors.

Now, in an EPSRC-funded project with Dr Alex Gegov, the researchers are developing a digital sensor system for powered wheelchairs, which will use artificial intelligence to learn how well a particular user can drive, and adapt the level of support offered accordingly.

By reducing the effort needed to drive, the devices will allow some people to use a wheelchair by themselves for the first time, and make driving and steering far easier for many others, Sanders said.

“The system can automatically adjust itself for the child that is driving the wheelchair,” he said. “So for example, if the system knows, or very quickly learns, that a particular child is blind and has very little spatial awareness, then it can adjust itself to assist them in the best possible way.”

The sensors will be connected to low-cost microcomputers, such as the widely available Raspberry Pi devices, which will be equipped with AI software.

The system will be capable of interpreting hand movements and tremors, and will take into account factors such as skill, tiredness, and recent driving performance, to determine how much influence on the motion of the wheelchair it needs to have at any given time.

At least three different AI systems will be used to suggest a course of action, such as turn left or stop, for example, and a Decision Making System (DMS) will decide which of these suggestions to follow, based on information from the sensors and the needs of the driver.

“We will have a decision making system that sits above the AI systems, to decide which opinion is more valuable, and what are the risks of one [suggestion] over another,” said Sanders.

The system, which will be developed with the Chailey Heritage Foundation, a specialist school in East Sussex, can be fitted to existing wheelchairs, to minimise costs.

It will be tested at the university before being trialled at Chailey Heritage Foundation late next year.