车辆路径问题的混沌神经网络解法

Chaotic neural network-based solution to vehicle routing problem

利用混沌神经网络在解组合优化问题时具有的随机性和确定性并存的优点,对一类随机需求服从泊松分布的车辆选径问题进行了求解,提出了一种混沌神经网络求解算法,并与平均场退火算法和模拟退火算法进行了比较。结果表明,该算法具有很强的避免陷入局部极小点的能力和较强的全局搜索能力,较大地提高了优化的时间性能和求解质量,是求解车辆选径问题的有效方法。

By making full use of coexistence of randomicity with deterministic property in a chaotic neural network, a novel solution to a combinatorial optimization problem was proposed. For a class of Vehicle Routing Problems （VRP） with Poisson--distributed stochastic demands, a model was first set up to solve the problem and the solution was then compared with those obtained by the existing mean field annealing approach and simulated annealing approach. Results from case studies showed that the proposed algorithm could avoid getting stuck in local minima and has better global--search capability. The proposed algorithm has greatly improved optimization time property and solution quality, and it was an effective method to solve Stochastic VRP problems.