| The active investigation of complex environmental parameters involves many computationally intensive operations, such as data collection, aggregation, dissemination and processing. The adoption of multi-agent systems appears as a very promising approach in the development of a new generation of intelligent autonomous robotic agents for environment monitoring. This thesis discusses a multi-agent system whose global goal is the minimization of entropy of an environment, based on a novel tree-in-motion-mapping method. The proposed method combines simplicity and speed of computation, along with low storage and communications requirements, ideally fitting a system composed of a finite number of mobile robotic agents, each possessing limited sensing, processing and communicating operational entities. The method is also extensible to include multiple factors for the determination of the environmental entropy, hence allowing measurands such as electromagnetic, thermal and optical energies to be aggregated in the entropy function. Simulation results demonstrate the efficiency of the method, compare the centralized and decentralized approaches, and present the savings, both in processing and communication times, when compared to existing methods. Experimental results are presented and analyzed to validate the method in a testbed environment. Finally, a formal specification of a generic multi-agent system intended for environment monitoring is presented, while a virtual prototyping tool intended for verifying diverse scenarios in simulation, before proceeding to implementation, is outlined. |
