This project is for £581,954 and is 80% funded by the EPSRC for £465,563.
Investigating the novel use of Shared Control, Sensors and Artificial Intelligence to create systems that will significantly and positively impact on the lives of powered-wheelchair users. People will be able to drive for longer and in some cases for the first time.
A new technique that continuously assesses ability will share control between drivers and intelligent sensors. The work will improve access to independent mobility and allow at least some self-initiated mobility even without the spatial awareness and neural ability that is usually required, so that even some blind children will be able to steer without needing helpers.
The research will develop technologies that will enable the next generation of assistive devices to provide natural control through enhanced and intelligent sensor feedback. Choices for a particular driver will be bespoke, based on their
characteristics and history.
Aim: Create new systems to improve mobility and quality of life for people with disabilities.
1: Create AI to interpret what a human user wants to do.
This will reduce effort and stress. New digital object-proximity-sensing and some input devices will be created and simple systems to interpret them will be investigated, starting with simple IF THEN systems. Similar methods have been used before but this initial work is necessary for the new research. Then new Fuzzy Systems will be created to interpret hand movements and tremors from among other involuntary movements. This will be the only attempt that has been made using this promising method. Achieving this would allow many more people to use powered wheelchairs.
2: Sensor fusion to interpret the environment.
This will reduce tiredness and collisions. A Rule Based System will be investigated first to generate revised instructions to correct for veer; the first time that this has been attempted. New digital object-proximity-sensing systems and effortreduction systems will be created, along with new input devices and AI systems to monitor them. That will improve mobility
and allow some disabled people to use powered wheelchairs for the first time.
3: Create a Decision-Making System (DMS).
This will compare outputs from the AI systems and suggest best possible course of action. This will be the only time that this sort of approach has been attempted with a powered wheelchair. Intelligent reasoning will provide confidence weightings to improve the DMS. Confidence weightings will allow the DMS to efficiently select output. A Case Based Reasoning (CBR) system will be investigated for this and compared with other methods. Reasoning will be revised to compare errors against search criteria and then to investigate whether further improvement is possible. The CBR will be extended for this to investigate if results can be improved.
4: Share control between the disabled driver and the new wheelchair system.
Automatically assess the ability of the wheelchair user and establish control gains for the sensor system and human driver by calculating a self-reliance factor depending on ability, tiredness, recent driving performance etc. An avoidance-factor will depend on obstacle proximity, a safety-factor will denote the ability of the driver and an assistance-factor will depend on time spent driving and tiredness. The sensor system will influence the motion of the wheelchair to compensate in those areas. This will be the first time that a wheelchair system has adapted to a disabled human user in real time.
5: Carry out experimental test trials to verify that mobility and quality of life is improved.
Experimental testing at UoP will verify that the systems function correctly before being released for user trials. User trials will take place at Chailey Heritage (CH) to verify that mobility and quality of life is improved for people with disabilities.
Results will be assessed, documented, reported, published and disseminated.
Research will focus on the novel use of sensors and inventing new shared control systems and artificial intelligence (AI) to significantly and positively impact on the lives of both current and potential powered-wheelchair users.
Recently developed sensors will be digitised and then used in novel ways with AI to assist people with driving a powered wheelchair. This will allow some people to use a wheelchair by themselves for the first time, and will make driving and
steering easier for many others. That will reduce the need for carers, improve health outcomes and give disabled people an opportunity for more independent mobility. For some it will provide mobility for the first time.
Access to independent mobility is important for self-esteem and a feeling of wellbeing. Natural independent mobility suchas crawling and walking are usually acquired in the first two years of life; if this does not happen then people can find it difficult to acquire the skills later. Currently a wheelchair can provide some self-initiated mobility but it cannot be introduced unless a person has the spatial awareness, physical ability and cognitive skills to understand the concept. Being able to transport oneself has a positive effect on general development that cannot be underestimated. This research will provide that opportunity.
Research at the University of Portsmouth has already resulted in analogue collision avoidance and effort-reduction systems, so that people can drive for longer. Work at the Chailey Heritage Foundation created track systems to guide
wheelchairs and novel systems that can follow a path parallel to a wall and sensors to safely detect the environment. All the devices will be redesigned as digital systems to connect them to expert systems for improved control. The new digital versions will interface to microcomputers. The new systems will interpret hand movements and tremors to improve control further. That will allow end-users to steer their powered wheelchairs without needing helpers and provide a greater sense of accomplishment and freedom, whilst simultaneously helping to reduce carer costs.
The abilities of the wheelchair user will be constantly assessed so that control gains can be automatically set for the sensor system and the human driver. This will be achieved by calculating a self-reliance factor depending on ability, tiredness, recent driving performance etc. An intelligent avoidance-factor will depend on obstacle proximity, a safety-factor will denote the ability of the driver and an assistance-factor will depend on time spent driving and tiredness. The sensor system will influence the motion of the wheelchair to compensate in those areas. This is the first time that this has been attempted.
Different AI systems will be used for different tasks to capitalise on their separate distinct strengths in diverse circumstances. An original hierarchy based upon the structure of Artificial Neural Networks will be used to integrate them.
At least three AI techniques will be used to select courses of action for a wheelchair and a new Decision Making System (DMS) will be created to determine a best course of action by considering and comparing the outputs from the different artificial systems and the requirements of the human user. Each system will provide a level of confidence for a potential course of action, for example turn left, stop etc. The DMS will determine the action to take.
This EPSRC project will produce both new devices and new ways of integrating devices into wheelchairs to ensure safe navigation and personalized assistance with general low cost but automatically adjustable solutions that make the systems bespoke and adaptable in real time. This will help to ensure users achieve maximum functionality. The devices can be added to existing wheelchairs, providing a cost-effective way of improving quality of life and independence.