Dynamic Reconfiguration of Real-Time Distributed Systems Based on Agents

Currently distributed systems are present in several applications in industry and research, such systems are restricted to time conditions. The presence of failures in these time critical systems requires decision making to avoid loss of deadlines, performance, instability or total suspension of the system. A strategy for achieving this is the dynamic reconfiguration of the structure of the system aimed at counteracting the effect of such failures. This article presents a proposal for dynamic and distributed reconfiguration considering each element of the distributed system as an agent with communication and cooperation capabilities to act joint with other agents to generate a reconfiguration action. These actions are the modification of sampling periods of sensors agents, decreased activity of the agents with fails and / or use of voting algorithms. For each type of system failure correspond a particular type of reconfiguration which is reached based on exchange of information between agents and the agreement they reached. The proposed dynamic reconfiguration scheme is implemented in the flight control of a prototype helicopter with two propellers through the inclusion in the closed loop control of a distributed system that uses a common communication media. The results of the reconfigured system response are obtained through numerical simulations carried out using Matlab and Simulink, as Real-Time simulation tool was used TrueTime.