Research On Navigation Of Mobile Robot Based On Kinect And Potential Field

Intelligent mobile robots(called a mobile robot in this follow-up paper) is a robot system which is able to move intelligently in the environment by preserving the current surroundings. Mobile robot navigation problem is quite important for the mobile robot,which focuses on a collision-free movement of the robot when it moves from a starting position to a target position based on certain evaluation criteria in the environment where obstacles exist. And one of the key tasks for mobile robot navigation is path planning which aims to find a feasible path between the starting point and the destination point.Situational awareness is the key technology for mobile robot path planning.Ambient information can be captured by the robot vision and Kinect is widely adapted as the eye of robot because of its low price and being able to provide a wealth of 2D or3 D information.In this paper, Kinect is utilized to capture the environment information. Combined with 3D point cloud obtained by Kinect, the distance information of the current environment can be obtained after the depth map obtained from Kinect being refined.Then potential field of the environment can be established according to the distance.Finally, robot is able to do path planning based on the potential field.The main work and contributions of this paper include:1. Designed a new path planning algorithm. First studied various technologies for path planning especially focused on the potential field method. Traditional potential field method is an effective static path planning method with the shortcomings including the local minima problem, GNRON problem and poor performance in the dynamic enviroment. Motivated by the idea of traditional potential field method, this paper came up with a new approach which is called as Shadow Potential Field Method which overcame both local minima problem and GNRON problem while being able to be applied in dynamic situations. Robot can avoid the dynamic obstacles with the reference of the behavior of the goal while keeping as close as possible toward the goal. Such method is especially suitable for the accompany rorot. Simulation results based on MATLAB showed the effectiveness of this new algorithm.2. Refined the depth map quickly and efficiently by introducing point cloud as guide information along with bilateral filter innovatively. Introduced the principle of Kinect and how it captures depth map and 3D point cloud data, and then analyzed the reasons for the information lack of the depth map and studied various algorithms to repair the depth map where holes exist. In this paper, 3D point cloud was adapted as the guidance information to refine the original depth map by extracting the plane component through 3D point cloud based on the efficient index data structures kd-tree when searching in the point cloud space. Bilateral filtering is also adapted. Simulation results based on Visual Studio, PCL, Open CV and Open NI showed the effectiveness of this approach.3. Designed the overall framework of the Kinect-based mobile accompany robot navigation system. Robots used Kinect to scan the current environment and obtained the depth maps and three-dimensional point cloud. Then, established the potential field based on the depth map by making most of three-dimensional point cloud. Finally, the robot was able to do path planning based on the established potential field.