动态优化策略广泛应用于很多领域,包括基于微分方程模型的最优控制问题求解等等。这类问题在离散后,得到大规模、复杂的非线性优化问题。现在的优化方法采用传统收敛准则,往往收敛速度缓慢甚至最终收敛失败,无法及时得到可靠的数值解。本文针对联立法离散后的问题,采用一种基于降精度求解准则(reducedprecision soluution criterion,RPSC)的序列2次规划方法(sequential quadratic programming,SQP)进行求解。RPSC定义了一系列指标,用于判断继续迭代是否能够有效改善解值,是否可以终止求解进程。仿真实验结果表明,该收敛准则与传统收敛准则相比,能够及时终止求解过程,同时返回较好的数值解。此外,收敛闽值可以根据用户的实际需要设定,在得到满足要求的估计值时及时终止求解进程。
Dynamic optimization strategy has been applied in many domains,like the numerical solving of optimal control problems(OCPs)based on ordinary differential equations(ODE)model et al.The OCP problem is transformed into large-scale nonlinear optimization problem after discretization.It is usually difficult or slow to converge when the OCP is solved by optimization algorithms based on traditional termination criterion. This paper presents a kind of novel termination criterion named reduced precision solution criterion(RPSC).RPSC defines a series of indices for judging whether current iterate is good approximation and whether the solution procedure should be terminated.The RPSC is integrated into sequential quadratic programming(SQP)algorithm to solve the OCP.The simulation results demonstrate that RPSC can terminate the OCP solution process quickly with satisfied results.