Research On Mobile Manipulator Of Environment Perception And Trajectory Planning

Mobile robots have found an increasingly wide utilization in recent years, this make a higher request for robots’ adaptive capacity to unknown environment. The uncertainty and complexity of environment require mobile robot better perception and planning ability to different occasions. So concentrate on the task of dangerous source searching and grasping, two basic theory well studied in this paper are path planning of a mobile platform based on discrete global environment model and the mobile manipulator environment perception and motion planning based on information entropy, this paper intends to achieve key steps of target object disposal such as large unknown environment construction and path planning, local obstacle environment real-time modeling and mobile manipulator trajectory planning.First of all, this paper builds a mobile manipulator’s environmental sensing system using RGB-D(Kinect V2) vision sensor, monocular vision sensor and 3D laser sensor. a global environment point cloud data is acquired by Kinect V2 that mounted on the back of the mobile platform, after it is processed to generate global discretization model which is used to guide the path planning; combining a 2D commercial laser scanner with a designed gimbals-mechanism, a 3D laser scanner is constructed, which is mounted on the end of mobile manipulator to obtain local 3D discrete environmental information, and it is equipped with monocular vision sensor to collect the pose information of the target object. Unknown area is estimated by the information of the three-dimensional laser data based on established Gaussian process estimation model, which will includes the environment geometry information entropy and hazards pose multidimensional discretization environment model, and information entropy is to provide reference information for subsequent mobile manipulator motion planning.Secondly, in the process of global environment exploring and modeling, the mobile platform obtains 3D point clouds and hazard information of multidimensional environment model through the environmental sensing system, this paper establishes an improved artificial potential field path planning method based on discretization of the continuous environment and a 2D grid map transform. Finally guide the mobile platform to reach the global target, which is suitable for mobile robot arm grab the location of the source of danger. This paper builds an environment perception and reconstruction based on ROS and mobile manipulator trajectory planning software system after detailed analysis of the task based mobile manipulator planning process. A trajectory planning algorithm of mobile robot arm based on Gauss process information entropy is put forward, and it can be used to realize the task of dangerous source disposal in unknown complex environment. Based on the probability and information entropy of the unknown point in 3D discretization environment model, the objective function of the end point trajectory of the mobile manipulator is established. The space posture of manipulator is obtained by inverse kinemics, which need collision detection using discrete environment model predicted by Gaussian process. Repeating the process until reaching target point, a collision free mobile manipulator trajectory is obtained, and then it completes grasping and disposal task with marked dangerous source.Finally, the platform construction and experimental study of the mobile manipulator exploring, perception and autonomous grasping system are completed. The experimental researches are carried out In view of the environment perception system, the manipulator’s motion planning, which the feasibility of each component. The marked target is put into a large scale of complex environment, different obstacle models are designed to study on the environment perception and grasping trajectory planning of the mobile manipulator. According to the global discretization of environmental information perceived by Kinect V2, the mobile manipulator gets close to the target by the use of improved artificial potential field method for path planning, then it can arrive target point by monocular vision camera acquisition to complete grasping through 3D laser scanner to complete the local environmental perception and guide the coordinated motion planning of mobile platform and manipulator. It verify the feasibility of the system that the experiment of searching and grasping of a marked target in a large range of unknown environment is realized.