中文摘要：

为克服标准粒子群算法在求解高维TSP问题时求解精度不高、易陷入局部最优等不足,将每个粒子均赋予质量和加速度,利用泊松分布和牛顿第二运动定律动态调整粒子加速度,并将粒子维数以相似度划分为优势部分和劣势部分,正常更新时只对劣势部分进行相应处理,保持并扩大其优势部分以提高收敛速度,扰动时更新其优势部分以达到远离当前粒子网络的目的来跳出局部最优。当有粒子碰撞时,引入反向学习策略处理粒子,选择合适的降速模型来提高收敛速度。最后,将改进后的算法用于求解TSPLIB中的标准实例问题,并与经典算法进行比较。试验结果表明,提出的新算法在求解旅行商问题时具有高效率、低迭代次数及强收敛等特性。该结果可为智能算法在求解优化问题时提高精确性和加快收敛等方面的研究提供一定的参考。

英文摘要：

In order to overcome the bad convergence and accuracy of the standard particle swarm optimization algorithm in solving the high dimensional TSP problem,each particle was given their own character,such as mass and acceleration was given,and Newton second lawwith a Poisson distribution was introduced to dynamic control the particle acceleration. In addition,the particle dimensions were divided into advantages and disadvantages section based on its similar to reduce the dimension of update,normally update would only change the disadvantage parts to keep and extend their advantage parts so that it could improve the convergence speed,when the disturbance it would update its advantage parts to away from the current network so that it could jump out of local optimum,when particles collide,opposition-based learning strategy was used to deal with disadvantage section,and a better model of slowconvergence was selected. Finally,via numerous simulations of TSPLIB and comparison with other classical algorithms,the results showed that the improved algorithm had the feature of high efficiency,lowcomputational complexity and strong convergence,which were especially crucial for the functioning of large-scale distribution problems. Research results could provide a reference for the study on intelligent algorithms in solving optimization problems,such as howto improve the accuracy and speed up the convergence.