Directed Exploration for Goal Oriented Navigation in unknown environments

Autonomously navigating a robot in an unknown environment filled with obstacles to a goal location using a directed exploration strategy is presented. The exploration is directed towards the goal by iteratively selecting intermediate target points in the environment that reduces the distance estimate to the goal from robot's current position, until the goal is reached. A repetitive rechecking approach is also introduced to quickly detect and recover from dead-ends present in the environment. The proposed method alleviates the need to have access to the grid cell representing the goal, at the start of the navigation mission, as required by many dynamic grid path planning algorithms such as D*. In addition, the proposed method inherently supports any mapping application that outputs a grid of mapped information unlike many dynamic planning methods that require additional changes to support particle filter based mapping systems due to map orientation changes. Simulation experiments conducted in multiple environment conditions reveal that the proposed method is comparable to the optimal performance of dynamic path planning algorithms while requiring lesser number of decision steps to reach the goal.