On Mobile Robot Local Motion Planning Based On Laser Scanner

Self-localization, environment detection and real-time motion planning are fundamental for mobile robot’s autonomy. Since laser scanner can provide high-speed and-accuracy information about environmental objects distribution, it has become an essential sensor for mobile robot systems. This thesis proposes a Player Layer based robotic simulation and development platform, and further studies the analysis and processing methods of laser scanner’s high performance range data, in order to excavate laser scanner’s potential to enhance robot’s autonomous functions such as environmental cognition and real-time motion planning.This thesis first introduces the thesis’s background, significance and recent development of autonomous robots, and discusses laser scanner’s applications in robot localization, environment detection and motion planning. Then robot’s hardware platform and its kinematics model, laser scanner’s operating principle and performance are presented in detail, together with the Player Project based robotic soft-framework and the multi-PC simulation and development environment.Secondly a robot localization method based on online identification of landmarks in indoor environment is proposed. In this method, laser scanner’s ranger data are first clustered by their relative distances, then the points’ distribution of the data clusters are further online examined to identify the landmark’s data clusters and locate the landmarks. With the landmarks, reference coordinates can be established as reference for the estimation and calculation of robot’s position and posture. Further by the conversion and fusion of robot’s ranger data in different positions and postures, information about environmental objects distribution can be obtained. Since this method relies on the on-line processing and identification of ranger data and doesn’t need to calibrate the landmarks in advance, it is more suitable for the self-localization of robots application in unknown environment.Since laser scanners can provide robot ranger data of a fan-shaped view of environment, a data processing method based on multi-expansions for laser scanner data is proposed. By expanding obstacle edges with several ratios and analyzing characteristics of expansion maps, the proposed method extracts shape features of local environment to enhance robots’ ability of local environmental recognition and obstacle avoidance. Compared to related research, by analyzing and extracting environmental shape features, the proposed method can avoid relative methods’loss of environmental details due to the numerical description of environment. Simulations certify that the proposed method is competent for robots’motion planning and navigation in different environments, and real platform experiment validates the method’s feasibility for practical applications with low-precision control signals, response delay and tire slippage.Finally, main results of the thesis are concluded and summarized, and possible directions of further extending of the work are discussed.