【作者】 辛江慧；

【导师】 李舜酩；

【作者基本信息】 南京航空航天大学 ， 车辆工程， 2009， 博士

【摘要】 自主行驶车辆在军事和民用高科技领域有着广泛应用前景,它的主要应用包括侦察、监视、目标搜索、爆炸物处理、安全巡逻等。本文以自主行驶探测车的路径规划与避障技术为核心内容,在参与设计和制造“EV-II”号探测车原型车的基础上,重点研究了障碍识别、路径规划算法、动态避障算法,并将这些算法运用到原型车的试验中,验证了本论文研究成果的正确性。论文的主要研究内容和成果包括:(1)系统论述了各类轮式移动机器人的发展状况及其特点,系统分析了轮式移动机器人的体系结构、多传感器信息融合、路径规划等技术的发展状况和不足之处。(2)自主设计了探测车软、硬件系统,并制造了“EV-II”号探测车原型车。包括探测车本体机构、驱动系统、控制系统和电源系统。基于DSP和PC104的开放式控制器搭建了控制系统硬件平台,并根据实际控制需要选取合适的操作系统;根据探测车的技术指标,计算驱动系统所需要的动力,根据动力因素选择了合适的驱动电机,并给出了电机调速的具体过程;根据探测车的控制系统和驱动系统需要设计了探测车的供电系统。(3)设计了探测车障碍识别硬件系统并提出了一种新的基于信息融合的障碍识别算法。为了保证探测车在地面上安全自主行走,对比各种距离传感器的优、缺点,设计了基于DSP的多超声波获取距离信息的电路,给出了多个超声波传感器所测距离数据和嵌入式计算机PC104的串口通信过程,并采用视觉远程辅助导航。提出了一种基于模糊贴近度的数据融合新算法,缩短探测车导航中环境测量与数据处理的时间,提高了采集数据的精度和效率。(4)基于粒子群优化算法研究了可视全局路径规划问题。综合利用人工势场法及粒子群优化算法的优点,针对静态环境已知的探测车路径优化问题,运用等分法进行环境建模,在粒子群优化算法的基础上,引入人工势场法的斥力场函数,提出一种基于路径长度和路径危险度的适应度函数,根据全局最优解和局部最优解自动在线调节学习因子,从而在初始化及更新过程中能自动避开障碍物,快速安全地实现全局静态路径规划。(5)提出了一种基于速度障碍、碰撞危险度概念的动态环境模型,结合新的模糊神经控制算法,实现探测车在复杂的动态环境中自主安全地行走。针对动态环境中难以解决的探测车路径规划问题,基于模糊神经控制算法改进控制器的输入/输出模式,新的算法增加了模式匹配和加权平均两个要点,去掉了繁琐的模糊化和精确化过程。仿真实验表明,在提出的基于碰撞危险度的神经模糊算法控制下,探测车能够自主避开复杂动态障碍物,并且采用优化策略使得探测车朝向目标点行走,且不会陷入陷阱中。(6)综合利用上述的研究结果,设计并实现了一个探测车路径规划系统,并在自行研制的探测车平台上进行了大量的实车试验,验证了上述方法与算法的可行性与合理性。更多还原

【Abstract】 Autonomous have many valuable attributes that can benefit human beings in all fields of modern life. The exploitation of space&ocean provides a huge market for robotics. The functions of autonomous vehicles include reconnaissance, surveillance, target acquisition and so on. Several important Autonomous Exploration Vehicles technologies are discussed in these topics which include: Path Planning, Obstacle Avoidance, Perception Technologies, Control System Architecture and etc. All the research work in this paper are not only discussed theoretically autonomous about the obstacle avoidance but also performed with the intelligent four-wheeled vehicle“EV-II”.The experiments’results of the Explorer under different conditions are presented. The main content and achievements are as follows:(1)The present state and the feature of the wheeled robot in the world are summarized.The Perception Technologies, the Multi-sensor data fusion, and the path planning technology are integrated systematically and comprehensive.(2)Exploration vehicle body, driven system, power system and control system are designed. The“EV-II”body is made. The open controller based on DSP and PC104 is built. According to request of the actual control, the appropriate operating system is selected. Based on technical indicators the required power to drive system is computed. According to dynamic factors, motors are selected and adjusted. The embedded hardware platform of the driven system is set up. According to the control system and power system, the power supply system of the driven system is designed.(3)The hardware of obstacle recognition is designed.The data fusion algorithm is supposed.In order to ensure the travling safety of the exploration vehicle on the ground, the advantages and disadvantages of a variety of sensors are compared, the circuit of multi- ultrasonics which to obtain distance information is designed. The communication process between the distance information and PC104 are given. The visual-aided navigation is adopted. A new algorithm based on fuzzy close-degree of data fusion is proposed. So the navigation in environmental measurements and data processing time are shorten. Data collection speed and accuracy is improved.(4)Based on particle swarm optimization algorithm, the visualization global path planning is studied.Aimming to the exploration vehicle’s path optimization problems in the known static environment, equal portions France is used to carry out environmental modeling. Based on the particle swarm optimization algorithm, the repulsion field function of artificial potential field method is introducted. a new path planning approach based on artificial potential field (APF) and particleswarm optimization (PSO) is presented. And the learning factors are adjusted automatically on-line according to the global optimal solution and local optimal solution. The first step is to make a danger degree map(DDM) based on the repulsive force of obstacles in the environment. Then the PSO whose fitness function is the weighted sum of the path length and the path danger degree is introduced to get a global optimized path. The method has a simple model and a rapid convergence which can meet the safe and real-time demands of robot navigation. The feasibility and effectiveness are proved by the simulation results.(5)In the dynamic environment, the problem of exploration vehicle dynamic path planning is difficult to solve。Then a mathematical model of dynamic environment is proposed based on velocity obstacle and the concept of risk degree for collision, also a method of path planning with improved fuzzy neural network is given. The input / output of the controller are considered into the precise directly. So the improved algorithm includes only two elements: pattern matching and weighted average, thus the tedious process of fuzzication and precision are removed. The concept of patterns and pattern matching are introducted. Patterns includes input mode and rules mode, the norm is used to express the matching degree between them. According to the matching degree, a weighted average algorithm is used to determine the output. The Simulation is carried in the environment of removed obstacles and encountered obstacles. The result shows that method is valid. The exploration vehicle could avoid the complex and dynamic obstacles, walk towards the target point using optimization strategy, and could not fall into the trap.(6)By applying the above results of the study synthetically,the controller based path planning system of exploration vehicle is designed and implemented. A large number of tests on the exploration vehicle platform is finished. The feasibility of the above methods and algorithms’reasonablity are verified by the tests.更多还原