The advancement of an obstacle avoidance bayesian neural network for an intelligent wheelchair

In this paper, an advanced obstacle avoidance system is developed for an intelligent wheelchair designed to support people with mobility impairments who also have visual, upper limb, or cognitive impairment. To avoid obstacles, immediate environment information is continuously updated with range data sampled by an on-board laser range finder URG-04LX. Then, the data is transformed to find the relevant information to the navigating process before being presented to a trained obstacle avoidance neural network which is optimized under the supervision of a Bayesian framework to find its structure and weight values. The experiment results showed that this method allows the wheelchair to avoid collisions while simultaneously navigating through an unknown environment in real-time. More importantly, this new approach significantly enhances the performance of the system to pass narrow openings such as door passing.