Achieving reliability using behavioural modules in a robotic assembly system

• Main Author: Wilson, Myra Scott
• Published: University of Edinburgh 1992
• Subjects: 621.3
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.663922

The research in this thesis looks at improving the reliability of robotic assembly while still retaining the flexibility to change the system to cope with different assemblies. The lack of a truly flexibly robotic assembly system presents a problem which current systems have yet to overcome. An experimental system has been designed and implemented to demonstrate the ideas presented in this work. Runs on this system have also been performed to test and assess the scheme which has been developed. The Behaviour-based SOMASS system looks at decomposing the task into modular units, called Behavioural Modules, which reliably perform the assembly task by using variation reducing strategies. The thesis work looks at expanding this framework to produce a system which relaxes the constraints of complete reliability within a Behavioural Module by embedding these in a reliable system architecture. This means that Behavioural Modules do not have to guarantee to successfully perform their given task but instead can perform it adequately, with occasional failures dealt with by the appropriate introduction of alternative actions. To do this, the concepts of <i>Exit States</i>, the <i>Ideal Execution Path</i>, and <i>Alternative Execution Paths</i> have been described. The Exit State of a Behavioural Module gives an indication of the control path which has actually been taken during its execution. This information, along with appropriate information available to the execution system (such as sensor and planner data), allows the Ideal Execution Path and Alternative Execution Paths to be defined. These show, respectively, the best control path through the system (as determined by the system designer) and alternatively control routes which can be taken when necessary.