Smooth anytime motion planning using Fast Marching

The mobile robots needs to calculate accurate paths in order to execute multiple tasks. An anytime path planning algorithm in unstructured environments using the Fast Marching Method is presented in this work. Anytime algorithms have been successfully applied to robotics races with real cars and other robotic tasks in order to quickly generate a first path and to improve it in an incremental way when time is available. When the mobile robot is in an unstructured environment, it is necessary to include real-time information about its surroundings to obtain safer paths. This information is acquired by the robot by using a laser range finder. One important characteristic that this type of algorithms must satisfy is that the path must be generated in real time. The proposed method works in two steps. First, a smooth and safe global path is generated. This path is divided in multiple subpaths separated by equidistant nodes (defined by topological or metric constraints). After that, the obstacles information is added and a different path is calculated only when the original path is unreachable. The recalculation of the path is made up to the next node, which reduces the computational time. Different tests have been carried out in an indoor environment. The most important advantage with respect to similar approaches is that sub-paths are always efficiently generated in one execution cycle in terms of smoothness and safeness. Besides, the computational cost is low enough to use the algorithm in real time.