Optimization Strategies For Mobile Manipulator Path Generation

The motion planning of mobile manipulator is a popular field of robot motion planning,which not only has the driving ability of mobile platform,but also has the flexible operation ability of manipulator,which is both mobile and maneuverable,and has a wide range of applications.However,as a complex machine control system,the coordinated control of the overall system is also essential and worthy of study,provided that the mobile platform and the manipulator can work independently.Moreover,although some efficient robot path planning algorithms such as the A~* algorithm and the artificial potential field method(APF)have emerged,the performance of these algorithms decreases dramatically as the degrees of freedom of the manipulator increase and the dimensionality of the configuration space rises,and dimensional disasters may occur.Subsequently,sampling-based path planning algorithms such as rapidly-exploring random tree(RRT)and probabilistic roadmap(PRM)emerged to solve the motion planning problem of high-latitude robots.However,the efficiency of these algorithms in motion planning for complex environments such as narrow passages still exist,and the long planning time and low success rate in planning for narrow passages are the problems that these algorithms are facing.The main objective of this paper is the overall coordination control and path planning of the mobile manipulator.The main content of the study includes the following aspects:(1)We optimize the problem of low efficiency in constructing undirected graphs in probabilistic roadmap.For the construction phase of the undirected graph in the probabilistic roadmap,we use the k-dimensional tree(KD-TREE)algorithm to improve the efficiency of its search for neighbors,and the addition of the KD-TREE index makes the undirected graph planning is more efficient.After that,we smoothed the final route with a Bezier curve to make the final route more consistent with the robot kinematics.(2)The overall kinematic model of the mobile arm is analyzed.To address the problem that the mobile platform and the robotic arm in the mobile robotic arm are in different workspaces,we map the various variables of the mobile platform and the robotic arm into the same coordinate space for description from the perspective of spatial mapping,and determine the overall kinematics of the system so that the overall mobile robotic arm can be planned uniformly in the same coordinate space.(3)We propose and introduce a guided sampling strategy to optimize the PRM algorithm.We propose a guided sampling strategy for the dilemma that mobile manipulator is difficult to plan in the environment of narrow passages.The strategy of advanced localization and intensive sampling of narrow aisles results in a more reasonable distribution of sampled configurations throughout the configuration space,which is extremely effective in improving the planning success rate and reducing the planning time required.(4)We studied the overall collision detection method of the mobile manipulator.In view of the problem that the overall mobile manipulator is too computationally intensive for collision detection in three dimensions,we adopt the method of projecting the robotic arm sampling configuration in a two-dimensional plane to reduce its computation and enable the mobile platform and the manipulator to perform collision detection in the same dimension.At the same time,we also ensure that the manipulator can avoid obstacles through trajectory planning and maintain a safe state when the mobile platform moves toward the end point under collision-free conditions.