Mobile Robot Path Planning Under Nonholonomic Constraints

With more and more applications in various areas, the nonholonomic mobile robot has become a focus in the field of robotics. This paper emphasizes on nonholonomic mobile robot's path planning from the aspects of theory and application, and a series of path planning algorithms are also given to adapt to different environment, which could make the mobile robot accomplish the path planning task with the ability of autonomous environment adapting.A global path planner based on grid point representation is proposed to work in the known environment. Not only is an improved distance transform algorithm used to generate the desired holonomic path, but also is the cubic spline interpolation method used to smooth it. Then the A* method is exploited to search the valid path in the nonholonomic roadmap which is constructed by state lattices composed of grid points. The multi-layer state lattices algorithm is presented for the special environment with various narrow passages, so that the number of state lattice nearby the obstacles increases dramatically except the free space.A path planner based upon optimization is proposed, which takes full account of different impacts including the distance and angle between the robot and the target, the mobile robot's velocity and obstacles. This path planning problem could be transformed to a discrete optimization problem by using of the navigation evaluation function. In order to determine both the range of the searching speed and optimal speed of obstacle avoidance, the dynamic window method and the genetic algorithm are used respectively.A hybrid path planner combining global optimization and quick obstacle avoidance is proposed in partially known environment. The global path planner based on improved distance transform algorithm can generate a series of sub-goal-points to the target which are far from the obstacles. According to the information of the laser scanning, the steering control method is used as the local path planner which could avoid the unknown obstacles promptly. Simulations and experiment results in complex indoor environments show the method's effectiveness and practicability.In the last part of this paper, the POMDP navigation is studied and discussed, especially for the state estimation and action selection in the decision process. This work is very critical for the success of the POMDP navigation in the future work.