Consensus Study Of Multi-agent System Based On Model Predictive Control

The intelligence of group, shown by the cooperation of multi-agents, can perform complicated tasks which an individual cannot do. The cooperative control of multi-agents investigates how a large number of individuals with simple function can perform complicated tasks or achieve cooperative group behaviors by distributed control. As an important research aspect of the multi-agent system (MAS), consensus control of multi-agents has received more and more attention since last decade. In a MAS,"consensus"means that some of the agents'states reach consensus based on local information.Model predictive control (MPC) is the receding horizon control based on predictive model. With the known topology of multi-agent system (MAS) and appropriate optimization objective, MPC can speed up the MAS consensus extremely, and have high robustness on disturbance. This thesis focused on consensus problem of discrete-time MAS, and introduced MPC in consensus algorithm, further applied this algorithm in switching topology system, consensus in constant reference state, and connectedness preservation. The main research work of this dissertation can be summarized as the following three aspects:1. A study on MAS consensus problem with model predictive control. Considering the constraint of system topology, a consensus framework based on MPC is introduced. Under this frame, proposed a new consensus algorithmμ-MPC which uses future states information. Comparing with other consensus algorithms,μ-MPC improved consensus performance dramatically. Particularly, for complete graph system, it can realize one-step consensus.2. Parameters configuration ofμ-MPC. Since MPC is based on heuristic optimization, the choices of its parameters are essential for consensus performance. This dissertation analyzed five main parameters ofμ-MPC: step sizeε, optimization horizon H P, control horizon H U, predict weight matrixQ and control weight matrixR . The one-step prediction is particularly discussed.3. Theμ-MPC algorithm is extended to some more practical situations such as switching topology system, consensus on constant reference state, kinetic constraint of agents and connectedness preservation. Simulation results show that for all these problems,μ-MPC can solve them well.