Journal of South China University of Technology (Natural Science Edition) ›› 2006, Vol. 34 ›› Issue (6): 39-43.

• Electronics, Communication & Automation Technology • Previous Articles     Next Articles

Optimal Feedback Scheduling Based on Anytime Algorithm for Model Predictive Controllers

Zhou Ping-fang  Xie Jian-ying  Deng Xiao-long   

  1. Dept.of Automation,Shanghai Jiaotong Univ.,Shanghai 200030,China
  • Received:2005-06-17 Online:2006-06-25 Published:2006-06-25
  • Contact: 周平方(1976-)，男，博士生，主要从事实时系统、计算机控制系统方面的研究 E-mail:zhoupf@sctu.edu.cn
  • About author:周平方(1976-)，男，博士生，主要从事实时系统、计算机控制系统方面的研究
  • Supported by:

    国家自然科学基金资助项目(50405017)

PDF

866

Abstract:

A model predictive controller can be implemented as a MPC(Model Predictive Control)task with the characteristics of the Anytime algorithm ,which allows the computation time to be traded for control performance.This paper presents an optimal feedback scheduling(FS·CBS)algorithm for a set of MPC tasks to maximize the global control performance subjected to limited processor time.Each MPC task is assigned with a CBS(Constant Bandwidth Server),whose reserved processor time is dynamically adjusted.The constraints in the FS-CBS algo-rithm guarantee the schedulability of the total task set and the stability of each component．Simulated results indi-cate that the optimal scheduling algorithm is robust against the variation of execution time of MPC tasks at runtime,and that it performs much better than the basic CBS algorithm.

Key words: real-time control system, fedback, scheduling, model predictive controller, constant bandwidth server