Online Discovery of AUV Control Policies To Overcome Thruster Failures In order to improve fault-tolerance of Autonomous Underwater Vehicles (AUVs) we investigate methods to increase their reliability and persistent autonomy. We propose a learning approach that is able to discover new control policies to overcome thruster failures as they happen. The proposed approach is a model-based direct policy search that learns on an on-board simulated model of the AUV. When a fault is detected and isolated, the model is adapted to a new condition. The learned fault-tolerant policy is able to navigate the AUV towards a specified target with minimum cost, because the approach generates an optimal trajectory. The video shows a set of real-world experiments of an Autonomous Underwater Vehicle (AUV) in a thruster failure scenario. The applied thruster layout of the AUV, Girona500, and the control architecture of the AUV including the controller level and the fault recovery level are described in the video. In the first experiment, the functional Girona500 AUV is navigating using its original control system and it is able to reach the desired target. The right thruster of the vehicle is broken in the second experiment and since the AUV is still using its original controller, it deviates from the target position. In the final experiment, the AUV is navigating using the remaining functional thrusters while executing the learned fault-tolerant control policy. The result shows that the robot overcomes the thruster failure and reaches the desired target. Unlike most existing methods which rely on the redundancy of thrusters, our approach is also applicable when the AUV becomes under-actuated in the presence of a fault. Citation: ------------ S. R. Ahmadzadeh, M. Leonetti, A. Carrera, M. Carreras, P. Kormushev, D. G. Caldwell, "Online Discovery of AUV Control Policies to Overcome Thruster Failures, " IEEE International Conference on Robotics and Automation, (ICRA 2014), Hong Kong, China, 2014. More info: ------------- http://www.ahmadzadeh.info/ http://kormushev.com

• published: 14 Mar 2014
• views: 19