【作者】 杨建华；

【导师】 许宏科；

【作者基本信息】 长安大学 ， 交通信息工程及控制， 2007， 硕士

【摘要】 遗传算法是模拟自然界遗传机制和生物进化论而成的一种随机搜索优化方法,由于其隐含并行性和较强的全局搜索特性,使其具有其他常规优化算法无法拥有的优点。然而,与经典的方法比较,遗传算法还是一门新兴的学科,无论是在其理论上还是实现方法上都有待进一步完善,只有对其不断的改进,才能更好地发挥遗传算法的性能和特点,使其更广泛的应用于工程实践。在对遗传算法的基本原理、基本要素以及理论基础等进行详细分析后,本文针对基本遗传算法在应用中存在的局限性,提出了相应的改进措施:(1)从遗传算法自身的角度出发,采用了小生境技术的遗传算法,结合精英保留策略、种群多样性保持方案、新的适应度值标定方式、改进的自适应交叉和变异率对基本遗传算法进行改进;(2)在遗传算法的搜索过程中融合局部搜索能力强的梯度法,构成混合遗传算法来提高运行效率和求解的质量。随着国民经济的不断增长,人民生活水平的不断提高,汽车进入寻常百姓家中业已成为现实,随之而来的城市交通问题则日益突现出来。因此,采用现代科学手段,研究一些智能化的方法来解决城市交通管理问题,就成为当务之急。为了缓解城市交通拥挤,本文在分析了城市道路单交叉路口交通流特性的基础上,首先建立了以车辆平均延误时间最短,以相位有效绿灯时间和饱和度为约束条件的非线性函数模型,利用混合遗传算法对模型进行求解,得到在固定周期下的最优配时方案。仿真结果表明获得了理想的效果,表现了混合遗传算法的优越性。其次,针对交叉路口的拥挤状况,建立了以控制周期内路口的总的排队长度最小为目标,以相位有效绿灯时间和周期时长为控制变量的交通信号优化控制模型,利用改进的遗传算法对模型进行多次仿真计算,结果表明本文的优化控制方法能够使控制周期内路口的总延误排队车辆数明显减少,同时也体现了改进的遗传算法在解的稳定性、最优性和收敛速度等方面都优于基本遗传算法。更多还原

【Abstract】 Genetic algorithm is a random searching and optimizing method which simulates natural descendiblity mechanism and biology evolution theory. This method has some advantages that other usual methods don’t have because of its twocharacters——implicit parallelism and global searching. But after all, geneticalgorithm is a newborn optimizing method and both its theory and its realization need to be improved. Only in this way, can genetic algorithm apply to the practice more effectively and widely.After basic theory , basic factors and theory base of genetic algorithm being introduced detailedly, aiming at limitations on application of the simple genetic algorithm, this thesis bring forwards some improving measures: (1) From genetic algorithm itself, the niche technique is adopted and it combines elitists reservation model, scheme of keeping population diversiform, mode of demarcating fitness, improved adaptive crossover and mutation rate to improve basic genetic algorithm. (2) During the period of searching of genetic algorithm, blending steepest descent method whose ability of local searching is strong, forming hybrid genetic algorithm to enhance efficiency of circulating and quality of computing.With the development of our national economy and the improving of the civilian living standard level, the car has been popular for every one. The problem of urban traffic is increasingly serious. So it is urgent for all of us to solve this problem by adopting modern scientific technology and intelligent method. In order to mitigate city traffic jam, the traffic flow characteristics of urban intersections were analyzed. firstly, a nonlinear function model of urban single—point intersections was established, in which shortest average delay of vehicles were taken as objectives, and phase effective green time, saturation degree were taken as restrictions, the objective function of the model was solved by hybrid genetic algorithm. Solved result indicates that obtaining perfect effect, which manifests the advantage of hybrid genetic algorithm. Secondly, aiming at crowd status of intersection, an urban intersection controlling model was established, in which shortest total queue length of vehicles were taken as objectives, phase effective green time and cycle time length were taken as controlling variables, the objective function of the model was solved by improved genetic algorithm. The result indicates that controlling methods of the paper can make delayed vehicles of intersection fewer than before. At the same time, it incarnates that improved genetic algorithm is more excellent than basic genetic algorithm on stability, optimization and convergence speed of results.更多还原