摘要
提出一种基于简化模型的DHP(Dual Heuristic Programming)方法的学习控制,避免了标准DHP方法需要被控对象的精确模型来求得对于状态和控制动作的Jacobian矩阵,而是利用简化过程对象模型获得近似Jacob ian矩阵,实现学习控制的需要.生化反应器定值控制的仿真结果表明,该方法加快了学习过程,并对更大范围的参数变化具有鲁棒性.
The standard DHP method needs accurate plant model to calculate the Jacobian matrix of state and control action, which is difficult to meet. A learning control strategy based on DHP（Dual Heuristic Programming） method of simplified model is proposed, which applies approximate Jacobian matrix to DHP training and thus relaxes this limitation. Simulation results of contrapose Bioreactor show that the proposed method can accelerate learning process and is robust to larger parameter changes.
出处
《控制与决策》
EI
CSCD
北大核心
2006年第10期1087-1091,共5页
Control and Decision
基金
国家自然科学基金项目(60575033)
