Trajectory planner for a humanoid robot passing through a door

In order to support us, humanoid robots are expected to perform various tasks in our living spaces, such as offices. With regard to this purpouse, robots have to move from one room to another one autonomously. Especially in offices, robots must open heavy and spring loaded doors to pass through them. In this paper we implemented a trajectory planner for a humanoid robot to pass through a swing door. The trajectory planner takes into account matters such as collision avoidance with obstacles, stability of the walking motion during the physical interaction with the door, and physical restrictions imposed by the structure of the humanoid robot. We use the Artificial Potential Field (APF) approach in a new way: the artificial forces are summed with the ones due to the manipulation of the door, and the potential profile changes depending on the position of the robot to reduce the likelihood of deadlocks, particulary high when the robot has to pass through narrow spaces such as a doorway. In order to generate a peculiar trajectory, respecting several constraints, we use a Genetic Algorithm (GA) to evolve the APF. We verify the feasibility of the so optimized trajectory on the dynamic simulator for the humanoid.