A General Approach to Health Monitoring & Fault Diagnosis of Unmanned Ground Vehicles

With the current increasing complexity of tasks that robots have to execute, the occurrence of faults is inevitable. Faults can occur in different components throughout the robotic system ranging from sensors, actuators, hardware or software components. Therefore, it is crucial for robot designers to consider health monitoring and fault diagnosis of these different components. Usually, faults are handled by adding redundant hardware components, in order for them to operate in case one of the components fails, or, by modelling specific components or behaviours mathematically or statistically and comparing them with the actual system. Adding redundant components is usually expensive and sometimes not feasible. Moreover, modelling of complex components is an exhaustive procedure, and in many cases, the models are not accurate enough to mimic the behaviour of the component. Therefore, a generalised method for fault diagnosis and health monitoring that takes into consideration different components can be helpful for robot designers. Another aspect of health monitoring and fault diagnosis of robots, is the standardisation of messages. Bearing in mind the range of different components available to robots, it would be helpful to have a standard method of representing the health and faults of these components to aid the user in having a comprehensive overview of the status of the robot. Looking into NATO Generic Vehicle Architecture (NGVA), where different standards are developed and deployed for vehicle components in order to achieve quicker upgrades and interoperability, a standard for usage and condition monitoring systems (UCMS) was developed but not deployed yet. In this study, a proposed system architecture for non-intrusive health monitoring and fault diagnosis of robots is presented, an overview of different fault diagnosis and health monitoring methods are discussed, and finally, the UCMS data model is evaluated to see if it can represent the health and faults of a robot thoroughly.