Frontier Based Goal Seeking for Robots in Unknown Environments

In this paper, the problem of goal seeking by robots in unknown environments is considered. A frontier based algorithm is proposed for finding a route to a goal in a fully unknown environment, where only the information about the goal region (GR), that is the region where the goal is most likely to be located, is available. The paper uses the concept of frontier cells, which are on the boundary between explored space and unexplored space. A “goal seeking index” is defined for each frontier cell and used to choose the best among them. Modification of the algorithm is proposed with altered choice of frontier cells when wall like obstacles are encountered or when the robot falls in a “trap” situation, to further reduce the number of moves toward the goal. The algorithm is tested extensively in computer simulations as well as in experiments and the results demonstrate that the algorithm effectively directs the robot to the goal and completes the search task in minimal number of moves. The solution to the problem of local minimum is also addressed, which helps in easy escape from a dead-end or dead-lock situation. It is shown that the proposed algorithm performs better than the state of the art agent centered search algorithm RTAA*.