A NOVEL ARTIFICIAL POTENTIAL FIELD-BASED REINFORCEMENT LEARNING FOR MOBILE ROBOTICS IN AMBIENT INTELLIGENCE

Mobile robots are relevant for ambient intelligence (AmI) and play an important role in the application of AmI. The mobile robot can turn a normal environment into an AmI. It is a new mid challenging issue to design an excellent mobile robot to achieve the above action. The key problem that, should be solved is path planning for die mobile robots In this paper, a new method is proposed. In this met,hod, reinforcement learning (RL) problem is first transferred to a path-planning problem using an artificial potential field (APF); then, a new APF algorithm is proposed to overcome the local minimum problem in the potential field methods with a virtual water-flow concept. The performance of this new method is tested by three well-known gridworld problems. Experimental results show the effectiveness of the method in this RL system.