Path-planning of mobile agent using Q-learning and real-time communication in an unfavourable situation

In this paper a scheme for multi-robot co-operation in a real time environment is presented which is undertaken using robot-to-robot communication through Data Terminating Equipments. Two mobile agents are considered in the workspace one of which is fully functional while the data acquiring functionalities of the other are impaired. The former is known as the Co-operating robot, while the later as Blind robot. The data acquisition tools like camera and IR sensors are embedded in the co-operating robot. On the contrary, the blind robot is devoid of any such data acquisition tools. Thus the co-operating robot, in motion, gathers the necessary information about its environment using Q-Learning. The camera embedded within the co-operating robot helps it to identify its kin i.e. the blind robot using image matching technique and thus helps to know the sate number of the later. After learning phase is completed the co-operating robot enters into planning phase and starts reaching the goal state using state-action pair table acquired from Q-learning. Suddenly, the co-operating robot faces some wheel problem and becomes stationary. The new approach of the paper lies in co-operating the blind robot and guiding it to the goal state of the cooperating robot as a representative of the later. The novelty of this work lies in the use of wireless communication for data transfer from one robot to the other through the data terminal to support twin robot co-operation. The above approach lies in conformation of the act of reduction in average uncovered target distance with the number of steps.