多智能体避障路径规划研究

发布时间：2018-03-27 16:59

本文选题：多智能体　切入点：不规则　出处：《南京信息工程大学》2017年硕士论文

【摘要】：随着社会的不断进步,科技的飞速发展,人工智能的重要性逐渐在科学研究与工程应用中体现。其中的多智能体技术更是独树一帜,理论和实践都自成一个体系。同时,智能体的智能性、环境的复杂性和多样性,为多智能体相关的系统进行避障路径规划的研究提供了新的挑战,亦为热门突破和创新点之一。如今,对于障碍物的研究已不仅仅只停留于有规律可循的规则形状,运动状态也从静态过渡到了动态。为解决普遍适应规则的制定这一难题,论文假定利用现代通信技术可以锁定跟踪目标,从路径规划角度出发,针对不规则障碍物,就如何实现多智能体避障路径规划展开了研究,内容主要包括:1、单个静态不规则障碍物避障路径规划。传统的多智能体避障算法在考虑障碍物形状时,存在路径冗余、能耗高等现象,不具普适性。为此,给出一种不规则障碍物避障路径规划的算法,定义了自动识别凸形化的方法,并在融入子登陆点法完成避障的同时,实现边界最短路径规划,提高了障碍物处理的适用性和避障路径规划的智能性。2、多个静态不规则障碍物。考虑到现实环境中障碍物的多样性,利用自动识别凸形化规则对多个不规则障碍物环境进行了数据的集中处理,并改进子登陆点的概念,同时增加了辅助寻迹线作出瞬时路径规划,最后为避免“摩擦”碰撞现象的产生,增加了安全阈值、转角速度及最小运动速度等参数的定义,选择以避障为优先完成目标追踪的路径规划。3、单个匀速直线运动的不规则障碍物。为实现动态的不规则障碍物避障算法的尝试,将使用自动识别凸形化规则之后的障碍物采用组合运动的方式实现匀速直线运动;同时,加入碰撞预测和边缘运动策略,从而保证避障的优先性,并重新进行路径的规划,体现了环境的动态性和规划的智能性。4、实地测试。为实现算法的理论仿真与实际运动路径的比较,进行了智能车的实地测试,通过两者之间的比较,验证了算法的可行性和有效性。
[Abstract]:With the continuous progress of society and the rapid development of science and technology, the importance of artificial intelligence is gradually reflected in scientific research and engineering applications. The intelligence of the agent, the complexity and diversity of the environment provide a new challenge for the research of obstacle avoidance path planning in multi-agent related systems, and also one of the hot breakthroughs and innovations. In order to solve the problem of making rules of universal adaptation, the study of obstacles has not only been in the regular shape, but also in the state of motion from static to dynamic. Based on the assumption that the tracking target can be locked by using modern communication technology, this paper studies how to realize obstacle avoidance path planning of multi-agent from the point of view of path planning and irregular obstacles. The main contents include: 1, single static irregular obstacle avoidance path planning. The traditional multi-agent obstacle avoidance algorithm has the phenomenon of path redundancy and high energy consumption when considering the shape of obstacle. In this paper, an algorithm of obstacle avoidance path planning for irregular obstacles is presented, and the method of automatically recognizing convexity is defined, and the shortest path planning of boundary is realized by incorporating the sub-landing point method to avoid obstacles. The applicability of obstacle processing and the intelligence of obstacle avoidance path planning are improved. 2. Multiple static irregular obstacles. Considering the diversity of obstacles in real environment, In this paper, the data of several irregular obstacle environments are processed by using the automatic recognition convexation rule, and the concept of sub-landing point is improved, and the instantaneous path planning of auxiliary tracing line is added. Finally, in order to avoid the phenomenon of "friction" collision, the definitions of safety threshold, angular velocity and minimum velocity are added. The path planning. 3, which takes obstacle avoidance as the priority to complete target tracking, is chosen, and a single irregular obstacle with uniform velocity linear motion is chosen. In order to realize the algorithm of obstacle avoidance of dynamic irregular obstacle, The obstacles after automatic recognition of convex rules will be combined to realize uniform linear motion. At the same time, collision prediction and edge motion strategies will be added to ensure the priority of obstacle avoidance, and path planning will be carried out again. It reflects the dynamic nature of the environment and the intelligence of planning. 4, field testing. In order to realize the comparison between the theoretical simulation of the algorithm and the actual movement path, the field test of the intelligent vehicle is carried out, and through the comparison between the two, The feasibility and effectiveness of the algorithm are verified.
【学位授予单位】：南京信息工程大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP18

【参考文献】