余泽泰 余盟 肖人彬 钟卫卫 王玉梅



摘要 为了提高狼群算法的适应能力,使其能在复杂环境下具有较高的寻优精度与速度,提出一种具有动态适应性的狼群算法。首先,提出一种动态分群算法,增强狼群算法的全局适应性;其次,定义一种差异度拟熵,构造自适应步长,提高算法在复杂环境下的精度与收敛速度;使用随机游走环节,增强算法的适应性,缩短计算耗时。在17种测试函数上与其他几种算法进行对比,显示所提出算法具有较好的精度与收敛速度。在三维无人机航线规划问题上验证了该算法的有效性与实用性。
关 键 词 狼群算法;动态适应性;分群;差异度拟熵
中图分类号 TP18 文献标志码 A
Dynamical adaptive wolf pack algorithm and its application
YU Zetai, YU Meng, XIAO Renbin, ZHONG Weiwei, WANG Yumei
(Artificial Intelligence and Automation, Huazhong University of Science and Technology, Wuhan, Hubei 430074, China)
Abstract In order to improve the adaptability of the wolf pack algorithm, and provide it with good performance of search accuracy and speed under complex environment, a dynamical adaptive wolf pack algorithm (DAWPA) is proposed. Firstly, a dynamic clustering algorithm is proposed to enhance the global adaptability. Secondly, a difference pseudo entropy is defined to construct self-adaptive variable step-size, which improves the search accuracy and convergence speed. Random scout method is used to enhance the adaptability and shorten the computing time. Then, DAWPA is applied on 17 benchmarks. Compared with several other algorithms, DAWPA has shown better performance on accuracy and speed. Finally, application of DAWPA on 3D UAV route planning problem has demonstrated its validation and utility.
Key words wolf pack algorithm; dynamical adaptability; clustering; difference pseudo entropy
群智能優化算法是通过模拟自然界中生物种群的行为方式,来求解复杂优化问题的智能算法。相比于传统的优化算法,群智能优化算法在解决一部分NP难问题上展现了良好的效果,得到了大量研究,在解决优化问题上得到了广泛应用。
国内外学者已经提出了多种群智能算法,如粒子群算法(Particle Swarm Optimization, PSO)和人工蜂群算法(Artificial Bee Colony, ABC)。这两种算法得到广泛应用,在优化中展现了独特的优势[1]。然而,面对复杂问题时,它们出现了较严重的早熟现象[2-3]。
为了解决复杂情况下的寻优问题,在狼群捕猎方式的启发下,吴虎胜等[4]提出了一种新的群智能优化算法——狼群算法(Wolf Pack Algorithm, WPA),并基于马尔科夫链证明了该算法的收敛性。由于WPA具有较好的全局搜索能力,自提出就引起国内外学者关注,并被应用于生产调度、图像处理、路径规划、云计算等问题[5],在高维问题上具有较好的收敛与鲁棒性[6]。……