Model-based Techniques for Dependable Decision Making in Groups of Agents Operating Autonomously
MetadataShow full item record
Massively distributed embedded systems are rapidly emerging as a key concept for many modern applications. Autonomous agents with mobile sensors are breakthrough concept in technology. However, providing efficient and scalable decision making capabilities to such systems is currently a significant challenge, especially to have flexible strategies with predictable performance in hard-to-predict conditions. My thesis first proposes a goal-oriented model to allow automated synthesis of distributed controllers, which implement and interact through models of different semantics. Scalability of descriptions is realized through defining the nature of interactions that can occur among decision modules while leaving to task of optimally implementing these interactions by the execution environment. Applications with data acquisition for CPS system are offered. The thesis also proposes an approach to performance predictive collaborative control of autonomous agents operating in environments with fixed targets. A trajectory generation algorithm which considers the physical characteristics of autonomous mobile agents, e.g., dimensions, weight, velocity constraints and many others. is used in modeling. An Integer Linear Programming based model is used to optimize collaboration to achieve maximum task accomplishment and flexibility. It also offers detailed experimental insight on the quality, scalability and computational complexity of the proposed method. Another important challenge for Cyber Physical Systems is data acquisition through groups of mobile agents to optimize monitoring. Each agent optimizes locally dose not necessarily result in overall optimization without global predictions. An asynchronous interaction scheme using gaming theory between agents to maximize the utility of the acquired data is purposed. Experiments study the effectiveness of the scheme in comprehensive data acquisition while minimizing redundant data collection.