Decentralized task allocation and task execution using autonomous agents cooperating in dynamic missions
The area of unmanned, autonomous aerial robotics remains a challenging research topic due to its growing complexity. There are two main components in successful completion of tasks. They are task planning and task execution. This thesis proposes a new decentralized system architecture by integrating two existing algorithms; a decentralized task planning algorithm called Consensus Based Bundled Algorithm (CBBA) with a behavior based task execution architecture called Subsumption Architecture (SA) we attempt to exploit the decentralized features provided by CBBA and SA both these architectures, and the behavior based reactive execution feature of subsumption architecture, which can be used in unknown or partially known territories. A further attempt is made in improving the performance of CBBA, by incorporating a relationship coupling logic. What is being proposed is that if the parent task is considered by an agent, it is most logical that the same agent performs the subtasks as well. This system is then integrated with SA. The system is able to dynamically allocate tasks and execute them. A case study for Intelligence Surveillance and Reconnaissance is designed, simulated and implemented.