Research On Vision-based Obstacle Avoiding For Mobile Robots

Vision-based Mobile Robot (VMR) is an intelligent mobile robot. It capturesimages from camera, processes the images and extracts the information of path andobstacles, and then dynamic plans a path from the start point to the goal withoutobstacles, to make it run autonomously and continuously on road or outdoor field.Its research involves theories and technologies of many disciplines, and embodiesthe latest achievements of information and machine intelligence. It has been moreand more widely used in many areas and has great research significance andapplication value. The main content of the path planning is to find an optimal orsub-optimal path from the start point to the goal, to steer the robot safely andbypass all the obstacles without collision on the way. In this paper, we mainly dosome research on the methods of dynamic obstacle avoidance for vision-basedmobile robotsFirst, we build an embedded platform of vision-base mobile robot in thelaboratory. It works with PC through wireless network. Detailed information ofcircuit and software architecture design is given here.Second, this paper gives the methods of road detection and obstacle extractionbased on the color feature information of the environment. Consider the largeamount of image data and time-consuming of the processing, we further extract thecharacteristic parameters of roads and obstacles, and map into the robot’smovement space coordinate system, to improve the system performance in realtime.Finally, more research is done on the algorithm of vision-based dynamicobstacle avoiding. For the traditional artificial potential field method has thedrawbacks of unnecessary collision avoidance and local minimum problems,respective method, a relative velocity based selective avoidance approach andbehavior based on the wall following, is taken. By drawing on the driver’s drivingexperience, one solution of dynamic obstacle avoidance control system based onfuzzy logic is proposed. Risk of collision with obstacles and the position relating to the robot are regard as input values, then by fuzzy reasoning, output the robot’ssteering angle and acceleration. Followed by computer simulation, effectiveness ofthese algorithms in complexity dynamic and static environment is fully illustrated.