Abstract
We have developed a task planning system for an autonomous robot to maintain equipment in hazardous environments such as nuclear power plants. Generally, maintenance tasks are semi-structured tasks, and a model-based robot system is used for their execution. However, it is difficult to make exact environment models constantly at a work site. Therefore, we have developed a practical task planning system which can perform desired tasks smoothly, even though environment models include some errors. In this system, state value named ‘sureness’ is introduced to indicate the amount of error in the environment models. By using this value, a better procedure can be selected from the task knowledge database for performing a task. In this paper, we describe a task planning method with sureness value for the autonomous maintenance robot system, propose a procedure to define sureness value using fuzzy mem-bership functions, and present a developed task planner for assembling a nut and a bolt to a flange based on above task planning framework. We verified that suitable robot control programs were generated from the operator's task command, and assembling task of a nut and a bolt to flange were performed autonomously by using the developed task planner.
