Low-cost Visual Tracking with an Intelligent Wheelchair for Innovative Assistive Care

This paper presents the development of a low-cost vision-based robotic wheelchair system towards autonomous convoying. The non-holonomic follower vehicle obtains visual real-time pose data of a known coplanar target installed on the back of the leading vehicle. This allows the tracking vehicle to mimic the path of the preceding vehicle, while maintaining a safe distance behind it with the aid of a controller based on the robot's kinematics constraints. A back-end visual filter is proposed in the planning strategy to overcome the noisy environmental information acquired from the camera as it tracks the vehicle in front. The effectiveness of the approach is evaluated in an indoor setting using data obtained from an instrumented wheelchair platform and a low-cost camera, and validated with observations from a laser range finder and derived (known) maps of the environment.