| Summary: | 碩士 === 國立臺灣師範大學 === 機電工程學系 === 103 === The main purpose of research is to improve the localization performance and result of path planning of wheeled mobile robot. In the area of localization, we solve the problems from localization, path tracking and robot kidnapped, and propose an improved Monte Carlo localization based on vector model. The specific improvements of the method are as follows: First, we include the discrete kinematic model of
wheeled mobile robot in prediction process. Second, use vector model instead of bitmap to solve the resolution problem. Third, not only tournament selection and elitism, but also variation mechanism are used in the resampling process. The extent of Variation is decided by distribution of particles. Last, the re-initialize process is added
into the algorithm, so that the system can be more robust and better to robot kidnapped problem. In the area of path planning, we include A* algorithm into vector model instead of traditional bitmap, and use the corners as nodes in A* algorithm to solve the limited of orientation of robot problem, and reduce the amount of turning points. To do this, we modify the process of A* algorithm and add the function that checks particles are moveable or not. Only pass particles can be considered into cost function. To avoid collision, we include the image erosion to remove the area which is too close to obstacles, so there is enough distance between robot and obstacles. In addition, path smoothing process is added into the algorithm to get smoother path after planning, so that the robot can pass through all nodes more efficiently. Overall, our research can let the localization algorithm be with more efficacy, and plan a path with good balance in shortest path and avoidance ability, so the robot can move quickly and stably to the goal.
|
|---|
