四足机器人环境感知、识别与领航员跟随算法研究
发布时间：2018-07-23 18:52
【摘要】：腿足式机器人在运动中只需离散的落脚点而不需要连续的路径,因此较轮式和履带式机器人其能跨越更为复杂崎岖的地形,具备更加优异的运动灵活性和环境适应性。按照机器人腿的数量可将腿足式机器人分为单足、双足、四足、六足以及多足机器人。其中四足机器人是模仿四足动物运动形式的一种机器人,它既有超于双足机器人的平稳性又避免了六足或多足机器人机构的复杂性,既能以静步态方式在复杂地形上缓慢行走又能以动步态方式实现高速行走。因而在腿足式机器人研究领域,综合其机构复杂度和稳定性,四足机器人是一种较优的选择方式。随着对四足仿生机器人研究的逐步发展,研究重点已由四足机器人的结构设计及运动稳定性分析过渡到如何提高其在复杂环境下的适应性,因而实现四足机器人对复杂环境的感知能力并提高其运动的自主性是当前一个主流的研究方向。本论文对四足机器人在复杂条件下的环境感知系统构建、地形识别、路径规划、领航员识别及跟随等内容进行了全面而详细的研究,主要内容如下：1、构建了三维激光扫描仪(或二维激光扫描仪+云台)+单目相机+TOF相机的四足机器人环境感知系统。该系统通过激光扫描仪与TOF相机共同获取环境的深度信息,其中激光扫描仪可进行大场景下较粗略的距离测量而TOF相机则可实现局部范围内较精细的地形信息获取。基于得到的环境深度信息,采用数字高程模型(DEM)进行地形描述,并通过计算各栅格的坡度、粗糙度、起伏度对地形进行识别.粗糙度由该栅格所处的坡度平面与其8邻域高程点的离散程度进行计算,避免了采用高程方差计算时对粗糙度的误检测。2、为了提高路径重规划的效率及机器人的安全性,该文对A*算法做了改进,并分别提出了增量式A*算法(IA*)、具有障碍物栅格扩展功能的扩展A*算法(EA*)及EA*算法的增量形式算法(IEA*)。IA*算法在初始时刻采用A*算法寻找一条最优路径,然后在其余时刻通过寻找当前路径与目标投影点的最优路径进行增量式路径规划。与A*算法相比IA*算法虽不能保证搜索到最短路径,但它不需要从起点重新进行路径规划,从而大大提高了规划效率。EA*算法通过增大栅格g值的方式实现对障碍物的扩展,这种方式可以使路径在远离障碍物的同时不会阻塞狭窄通道。3、四足机器人对领航员的准确识别是实现机器人领航员跟随功能的前提,为保证识别的稳定性,该文采用领航员佩戴特殊反光材料的方式实现其与环境特征的区分。四足机器人通过激光扫描获得的反射光强度信息对该反光标识进行探测,进而实现对领航员的识别与定位。同时通过坐标映射则可实现激光扫描仪测得的领航员位置在图像坐标系下的投影,从而可进一步采用图像处理的方式对领航员位置做进一步判断。机器人与领航员之间的路径通过IEA*算法进行规划。4、为了实现对四足机器人的运动控制,该文对四足机器人的运动学模型进行分析,通过该模型可得出机器人足端坐标与关节旋转角度的关系,然后通过设计的四足机器人单腿液压伺服控制器实现对机器人各关节作动器的伺服控制,进而实现了机器人的行走功能。四足机器人在路径跟随时的难点是控制其转弯特性,因而该文同时对四足机器人的转向运动进行了分析,并通过调整运动控制策略实现了对路径的跟随。
[Abstract]:Leg foot robots only need discrete foothold in motion and do not need continuous paths. Therefore, the wheeled and tracked robots can span more complex rugged terrain and have more excellent flexibility and environmental adaptability. According to the number of robot legs, the leglegged robots can be divided into single foot, bipedal, quadruped, and six feet. The quadruped robot is a kind of robot which imitates the form of quadruped. It not only has the stability of the bipedal robot, but also avoids the complexity of the six foot or multi foot robot mechanism. It can walk slowly in the complex terrain in the static gait mode and can walk in the dynamic gait way. With the complexity and stability of the leg foot robot, the quadruped robot is a better choice. With the development of the research on the quadruped bionic robot, the focus of the research has been transferred from the structure design and the motion stability analysis to how to improve its adaptability in the complex environment. It is a mainstream research direction to realize the ability of the quadruped robot to perceive the complex environment and improve the autonomy of its movement. This paper makes a comprehensive and detailed study on the construction of the environment perception system, terrain recognition, path planning, navigator identification and following capacity in the complex conditions of quadruped robots. The main contents are as follows 1, 1, a four legged robot environment perception system is constructed with a three-dimensional laser scanner (or a two-dimensional laser scanner + cloud platform) and a single camera +TOF camera. The system uses a laser scanner and a TOF camera to obtain the depth information of the environment, in which the laser scanner can measure a rough distance in a large scene and the TOF camera can be realized. Based on the acquired depth information, a digital elevation model (DEM) is used to describe the terrain, and the terrain is identified by calculating the gradient, roughness and undulating degree of the grid. The roughness is calculated from the level of the gradient of the grid and its 8 neighborhood elevation points. In order to improve the efficiency of the path replanning and the security of the robot, the A* algorithm is improved in order to improve the efficiency of path replanning and the security of the robot. The incremental A* algorithm (IA*), the extended A* algorithm (EA*) with the barrier extension function and the incremental formal algorithm of EA* algorithm (IEA*).IA* algorithm are used in the.2. In the initial time, the A* algorithm is used to find an optimal path, and then the optimal path of the current path and the target projection point is searched for the incremental path planning at the rest of the time. Compared with the A* algorithm, the IA* algorithm can not guarantee the shortest path, but it does not need to rearrange the path planning from the starting point, thus greatly improving the planning efficiency. The rate.EA* algorithm can expand the barrier by increasing the grid g value. This way, the path can not block the narrow channel.3 while the path is far away from the obstacle. The accurate identification of the four legged robot to the pilot is the premise to realize the following function of the robot navigator. In order to ensure the stability of the identification, the passage is worn by the navigator. The special reflective material distinguishes it from the environmental characteristics. The reflective light intensity information obtained by the laser scanning is detected by the quadruped robot, and the identification and location of the navigator are realized. At the same time, the position of the navigator measured by the laser scanner can be realized in the image coordinate system by the coordinate mapping. The path between the robot and the navigator can be further judged by the way of image processing. The path between the robot and the navigator is planned by the IEA* algorithm. In order to control the motion of the quadruped robot, the kinematics model of the quadruped robot is analyzed. The model can be used to get the robot foot sitting. The relationship between the standard and the rotation angle of the joint, and then the servo control of the robot joint actuator by the designed single leg hydraulic servo controller of the quadruped robot, and then the walking function of the robot is realized. The difficulty of the four legged robots in the path and at any time is to control the turning characteristics, so the paper also transfers the four legged robots at the same time. The movement is analyzed, and the path following is achieved by adjusting the motion control strategy.
【学位授予单位】：山东大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP242
，
本文编号：2140333
[Abstract]:Leg foot robots only need discrete foothold in motion and do not need continuous paths. Therefore, the wheeled and tracked robots can span more complex rugged terrain and have more excellent flexibility and environmental adaptability. According to the number of robot legs, the leglegged robots can be divided into single foot, bipedal, quadruped, and six feet. The quadruped robot is a kind of robot which imitates the form of quadruped. It not only has the stability of the bipedal robot, but also avoids the complexity of the six foot or multi foot robot mechanism. It can walk slowly in the complex terrain in the static gait mode and can walk in the dynamic gait way. With the complexity and stability of the leg foot robot, the quadruped robot is a better choice. With the development of the research on the quadruped bionic robot, the focus of the research has been transferred from the structure design and the motion stability analysis to how to improve its adaptability in the complex environment. It is a mainstream research direction to realize the ability of the quadruped robot to perceive the complex environment and improve the autonomy of its movement. This paper makes a comprehensive and detailed study on the construction of the environment perception system, terrain recognition, path planning, navigator identification and following capacity in the complex conditions of quadruped robots. The main contents are as follows 1, 1, a four legged robot environment perception system is constructed with a three-dimensional laser scanner (or a two-dimensional laser scanner + cloud platform) and a single camera +TOF camera. The system uses a laser scanner and a TOF camera to obtain the depth information of the environment, in which the laser scanner can measure a rough distance in a large scene and the TOF camera can be realized. Based on the acquired depth information, a digital elevation model (DEM) is used to describe the terrain, and the terrain is identified by calculating the gradient, roughness and undulating degree of the grid. The roughness is calculated from the level of the gradient of the grid and its 8 neighborhood elevation points. In order to improve the efficiency of the path replanning and the security of the robot, the A* algorithm is improved in order to improve the efficiency of path replanning and the security of the robot. The incremental A* algorithm (IA*), the extended A* algorithm (EA*) with the barrier extension function and the incremental formal algorithm of EA* algorithm (IEA*).IA* algorithm are used in the.2. In the initial time, the A* algorithm is used to find an optimal path, and then the optimal path of the current path and the target projection point is searched for the incremental path planning at the rest of the time. Compared with the A* algorithm, the IA* algorithm can not guarantee the shortest path, but it does not need to rearrange the path planning from the starting point, thus greatly improving the planning efficiency. The rate.EA* algorithm can expand the barrier by increasing the grid g value. This way, the path can not block the narrow channel.3 while the path is far away from the obstacle. The accurate identification of the four legged robot to the pilot is the premise to realize the following function of the robot navigator. In order to ensure the stability of the identification, the passage is worn by the navigator. The special reflective material distinguishes it from the environmental characteristics. The reflective light intensity information obtained by the laser scanning is detected by the quadruped robot, and the identification and location of the navigator are realized. At the same time, the position of the navigator measured by the laser scanner can be realized in the image coordinate system by the coordinate mapping. The path between the robot and the navigator can be further judged by the way of image processing. The path between the robot and the navigator is planned by the IEA* algorithm. In order to control the motion of the quadruped robot, the kinematics model of the quadruped robot is analyzed. The model can be used to get the robot foot sitting. The relationship between the standard and the rotation angle of the joint, and then the servo control of the robot joint actuator by the designed single leg hydraulic servo controller of the quadruped robot, and then the walking function of the robot is realized. The difficulty of the four legged robots in the path and at any time is to control the turning characteristics, so the paper also transfers the four legged robots at the same time. The movement is analyzed, and the path following is achieved by adjusting the motion control strategy.
【学位授予单位】：山东大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP242
，
本文编号：2140333
本文链接：https://www.wllwen.com/shoufeilunwen/xxkjbs/2140333.html
