The robot motion involves mainly three modules:
rflex that manages the lower servo-control loop, trans- mitting the reference speeds at the micro-controller.
ndd integrates a local avoidance procedure based on an algebraic instance of Nearness Diagrams. The input obstacles are provided by the aspect map.
vstp is a Very Simple [but very efficient] Trajectory Planner based on an algebraic visibility graph opti- mized with hash tables. A main visibility graph is pre- computed for the static segments of the map. Dynamic obstacles can be added and removed in real-time upon supervisor requests.
The strategy used to coordinate the implied modules is dynamically established by the supervisor. The objective is of course to reach the target zone while avoiding obstacles. The planned trajectory is an Ariadne's thread for ndd: the vertices of the broken line are sub-goals. Usually the supervisor has to intervene only if ndd does not make progress towards the goal. In such case, various strategies can be applied: computing of a new trajectory taking into account the encountered obstacles, waiting for a while, starting an interaction with people around, etc. The motion is over when the robot is inside the target zone.The maximum speed that the robot can achieve in this mode is about 0.6 meters per second.