Autonomy

BT_example

Autonomy is an important aspect of planetary robotics. In the SEAR project it was a design driver in order to compete at the SpaceBot Cup. The autonomy system of SEAR is designed hierarchically. The top level is represented by the task management and focuses on the decision making process. Other autonomous processes are implemented within the subsystem software, such as the grasping process or the autonomous navigation. The autonomy of SEAR is methodical, because it is strongly connected to the mission scenario for which is was created. The system cannot – and is not supposed to – operate autonomously under any condition. It is even required to wait for operator input in case undefined states are reached. The autonomy approach for SEAR’s task management system makes use of behaviour trees. Behaviour trees are based in the computer game industry and used in order to create artificial intelligence for non-player characters. They offer unique and dynamical possibilities to create complex behaviours flexibly. Most importantly, it is possible to designan autonomy in a very defined way, enabling an engineer to predict the robot’s decision very accurately. Behaviour trees consist of different logic blocks whichthemselves represent behaviour trees also.