An Investigation Into Key Issues Of Obstacle Avoidence Of AGV Based On Depth Camera And ARM-arch SoC

Vehicle automatic navigation is a kind of through the sensor to detect the terrain,realize the automatic machine of directional movement in the physical space vehicles.AGV obstacle avoidance is one of the research emphasis in this field,its main content is in the actual production environment,AGV can according to the specific obstacle avoidance strategy,ensure the vehicle run from the starting point to run to the end without collision.AGV obstacle avoidance method is a core technology of obstacle recognition: the detection and analysis of space environment,locating obstacles in information environment.Environmental detection is the precondition of AGV is used to identify the obstacles,the robot can obtain real-time environmental data by computer vision.This paper adopts the Kinect as the main sensors,use the gradient detection and DBSCAN clustering algorithm analysis obstacles in the environment,then the depth data mapped to the grid diagram,finally uses the D* pathfinding algorithm to complete the AGV path planning.Main content of this paper are as follows:1.According to the particularity of AGV vehicle production environment,the ARM architecture Raspberry Pi as main controller of AGV.Due to the lack of on ARM platform device camera of device drivers,independently writing the first Kinect ARM-arch driver.2.Introduction of gradient detection algorithm as a ground plane detection means,combined with depth map gradient detection algorithm and DBSCAN clustering algorithm,analytical depth data more quickly.First elaborated Kienct and the principle of the depth map is generated,and then puts forward the concept of depth image gradient,with ground plane and space relationships between objects as reference information,using the gradient detection algorithm to distinguish the range areas and obstacles.3.The depth of the design based on the device camera and ARM architecture raspberries pie main controller of AGV system's overall framework.AGV using Kienct detecting operating environment,obtain the distribution of data generated depth map.And by using the gradient detection and clustering algorithm to deal with the depth of the figure,generated description of AGV and obstacle position relationship between grid graph,finally using the heuristic D* path planning method,AGV dynamic path planning.