Fault Tolerant Deployment Search And Task Cooperative Contol Of Robot Sensor Networks PDF Download

This thesis focuses on developing of a distributed, efficient and fault tolerant multiresolutional architecture for sensor networks. For demonstrative purpose, a powerful simulation environment using 3D environment model has been developed. The robot network is composed of autonomous robots capable of working cooperatively equipped with single typed simple sensor. The developed layered control architecture is hybrid including both subsumption and motor schema control strategies. In this proposed control method, behaviors in different or in same layer are coordinated with an evaluator unit that overcomes the difficulties of subsumption based architectures in terms of behavioral coordination. The final coordination between these layers is achieved cooperatively. We performed many simulation experiments to test robot deployment, search and task execution. It is shown that some important parameters such as target reaching time, energy consumption, and communication range can be optimized if an approximate prior information about the environment is known. Robots executes task based on a task allocation algorithm. Market based auction method is used as a task allocation algorithm with completely different robot fitness evaluation method allowing a distributive problem solving. Six non-linear fitness functions are developed to increase the fairness, and fault tolerance of task allocation. These functions have been tested to represent the successes and failures of robots in a compact form. Performance analyses test results have shown that fairness increases two times more in task allocation when these fitness functions are used, compared to the results existing fitness evaluation methods used in the market based auction algorithms. Moreover, fault tolerance is increased by using fitness functions devoted to failure conditions.