Executive Summary : | Around 75 million people, or 1% of the world population, require wheelchairs daily, primarily for neurological and spinal cord problems, cognitive or visual impairments, and older individuals in geriatric care centers and airports. Autonomous wheelchairs, equipped with mobile robotic technology, face challenges in navigating socially acceptable ways that ensure physical safety and follow social constraints. Existing navigation techniques suffer from modelling uncertainties, large data requirements, long learning phases, and lack of safety guarantees. This leads to freezing problems, where all recommended velocities lead to collisions, causing discomfort or averting the wheelchair from reaching its destination. This project aims to develop a safety-certified socially aware navigation scheme for autonomous wheelchairs. The project proposes an Empathic Social Force Model for modeling crowd motion in the vicinity of the wheelchair, using socio-spatiotemporal characteristics for empathic motion prediction. A safe reinforcement learning technique will be developed for motion planning, providing safety guarantees through control barrier functions. The technique will be trained and tested in a crowded Indian environment, with the deliverable being a product for socially compliant navigation. The project aims to develop wheelchairs as assistive devices to support mobility-impaired individuals, empowering them with better accessibility, dignity, and equality. |