Self Detection of Failure in Robots

Abstract

Robotics is an emerging field and robots may be employed in places where human intervention is not possible. Multiple robots may work in a coordinated manner to achieve certain tasks. However one of the big problems is the detection and recovery from failures, since human intervention may not be possible. To this end we propose an autonomic self-detection and self-recovery robotics architecture based on the human immune system. In particular we look at two types of communication failure, failures caused by robot isolation and failures caused by intermittent message loss. This thesis focuses on one component of an autonomic robotic architecture namely, self failure detection mechanism in robots. Our goal is to make the robot recognize the failure encountered during its operation and send related failure information to the activation unit for further action. We propose an approach to self-detection based on observation graphs. Simulation results show that the failures were effectively detected. The proposed recognition unit is similar to T-cells in the human immune system.