Detection Of Road Conditions And 3D Map Building For Walking-Assistant Robot Based On Laser Range Finder

As China has entered the aging society, walking-assistant robot is becoming the focus of researchers around the world. As a service robot, there is of great need for the walking-assistant robot to help the elders or the disabled to walk, with some extra function. However, some conventional walking-assistant robots are pure mechanical products,and there are no extra functions except assistance, to say nothing of avoiding obstacles, and some walking-assistant robots use ultrasonic sensor to detect obstacles. However, the detection range of this sensor is quite limited.In addition, walking-assistant robots usually work in indoor environment. In the unstructured environment, walking-assistant robots should have the function to detect the road conditions, so if the road is uneven or obstacles arises, walking-assistant robot should remind the user to avoid the obstacles.What's more, since the users of walking-assistant robots are ordinary operators, who don't have professional skills to manipulate the robots. And the robot should have the function to localize itself and navigate so as to serve the human being. In this case, the map and manipulation of the robot must be easy to learn and the robot should establish a 3D map automatically.Therefore, the detection of the road conditions and building 3D map for walking-assistant robot is quite essential. This thesis makes use of the laser range finder to detect the road conditions and build up a 3D VRML map. Firstly, This thesis proposes a new function for walking-assistant robot to detect the road conditions and obstacles, as well as avoiding the obstacle automatically. In addition, as to the 3D map building, this paper sets up a platform to acquire the 3D point clouds of the indoor environment. and then uses ICP algorithm the register the point cloud in the environment, afterwards RANSAC algorithm and data clustering methods are used to segment the large-scale unorganized point clouds. Finally, the segmented point clouds are converted to VRML model and a 3D VRML map is built. The whole system is verified in different environments and the experimental results are in expectation.