Introduction
A typical robot navigation problem involves specifying a goal position, planning an obstacle-free trajectory from the robot’s current position to the goal position based on either a prior map (RRT, RRT*, A*, etc) or based solely on the robot’s sensor readings (Link) and getting the robot to follow the planned trajectory. In following the planned trajectory, it’s likely that the robot’s environment would change as dynamic obstacles like people, or previously unnoticed objects are introduced into the environment. As such the planned trajectory could end up being obstructed by obstacles. Generating a whole new plan in such cases isn’t always feasible since the planning process could be quite expensive. A more practical approach is to develop contingency behaviors the robot can fall back on to avoid the temporal obstacles and recover its way back to the original planned trajectory. The process of coming up with these contingency behaviors is called local planning.
[TO BE CONTINUED]