一类多智能体系统的渐近最优分散控制.pdfVIP

Proceedings of the 25th Chinese Control Conference 7–11 August, 2006, Harbin, Heilongjiang , , E-mail: litao@amss.ac.cn 100080 : Nash- , (LPDMAS) , . : , , , Nash- , Asymptotically Optimal Decentralized Control for a Class of Multi-Agent Systems Li Tao, Zhang Ji-Feng Academy of Mathematics and Systems Science, Chinese Academy of Sciences, Beijing 100080 E-mail: litao@amss.ac.cn Abstract: The concept of asymptotic Nash-equilibria with respect to stochastic performance index is introduced, a decentralized control law is constructed for Large Population Dynamic Multi-Agent Systems (LPDMAS) with costcoupled stochastic preformance indexes by using the state aggregation method. By the limit probability theory, it is shown that the closed-loop system is almost surely uniformly stable, and the control law is asymptotically optimal. Key Words: Multi-Agent Systems, decentralized control, stochastic differential game, asymptotic Nash-equilibria, stochastic performance index 1 (Introduction) , , . : (LPDMAS). ., ([2]); LPDMAS Agent ([5]) . Agent Agent Agent ([1]), Swarm ([3]); ([4]) , , . Agent , , IEEE Catalog Number: 06EX1310 , : . Agent () ,, , Agent , Agent . , Agent . LPDMAS Agent .Agent . , Agent , .Agent , , Agent , , , , . 346 Agent ; ; Agent ,, Agent . Agent , LPDMAS, , Agent ( ). , LPDMAS ,“ ” , “” Agent , Agent . , Agent ,, . LPDMAS , . [6]-[9] Agent , , Swarm, LPDMAS ,. LPDMAS , . [10] CDMA ;[11] ,, Agent , LQG . [11] , LPDMAS , Nash- , Agent ., : 1) Agent 2) Agent , 3) . , 4) . L2 , .· 2∞} Cn F- Cn [0, ∞) n Cnb = {x | x ∈ Cn, supt≥0 x(t) ; x ∈ Cnb , x C = supt≥0 x(t) ; {Wi(t), i ≥ 1} (?, F, P ) d ; L(?, F, P ) (?, F, P ) . 2 Nash (Problem Formulation and Nash Asymptotically Optimal Decentralized Control) N Agent SN , i Agent dxNi = [fiN (xiN , xN?i, t) + bNi uiN ]dt + g(xi, t)dWi(t), xNi ∈ Rn i Agent ,xN?i = [x1N , ..., xiN?1, xNi+1, ..., xNN ], uiN i Agent , UlN,i = {uiN | uNi σ(xiN (0), Wi(s), s ≤ t) } uNi , UgN,i , UgN,i , UgN,i = {uiN | uiN