Robotic fault tolerance: algorithms and architectures

Abstract

Fault tolerance is an essential factor in ensuring successful autonomous systems, especially for robots working in remote or hazardous environments. To avoid the cost and risk involved in sending humans into these environs and improve the chances of mission success, robots must be able to quickly detect and recover from internal failures without relying on immediate repairs or human intervention. In developing robotic fault tolerance algorithms, the possible failures within the robot and the interdependence of these failures must first be determined. One useful tool for performing these tasks is Fault Tree Analysis. The tree structures developed by this technique provide a flow chart of possible robot fault events and define the cause and effect relationships between the failures.