Mobile robots are widely used in many applications in industrial fields as well as in academic and research fields. The robot path planning problem is a key problem in making truly autonomous robots. It is one of the most important aspects in mobile robot research and plays a major role in their applications but is a complex problem. The role of path planning of mobile robot can be described as finding a collision free path in a working environmentenriched with obstacles from a specified starting point to a desired destination position called the goal. Additional characteristic of path planning in known environments is to satisfy some certain optimization criteria. Most of the traditional path planning approaches such as visibility graphs, cell decomposition, voronoi diagram, etc. are designed and functioning well in static known environments. However the real environments are consisting of both the stationary and moving obstacles. Artificial potential field based methods can be applicable for both the static and dynamic environments as well as for the known or unknownenvironments. Because of the analytical complexity of the dynamic environments, researchon path planning in dynamic environments is limited but there are hundreds of research workhave been reported on path planning in static known environments.Because of the mathematical simplicity, easy implementation and real time applicability ofartificial potential field, it has become popular in robot path planning. However, the potential field based path planning shows some inherit shortcomings such as dead-lock. Recently, in the field mobile robotics, some different techniques have been proposed to overcome the dead-lock issue associated with the artificial potential field based path planning. Most of these research work targeted only a specified situation where the dead-lock can happen.In this research, we proposed a method for avoiding robot from dead-lock caused in differentsituations of mobile robot path planning using artificial potential field. In the proposedmethod we have introduced a new repulsive force component which is depended on therobot’s heading direction. The proposed method is evaluated for different conditions whichcreate dead-lock for traditional artificial potential field method. The simulations of theproposed approach have indicated that it has a capability of avoiding dead-locking associatedwith the traditional method, and is simpler and easier to implement. However in realimplementation it is required to extract the geometric features of the environment such aswalls and corners for our consideration in structured environments. Consequently, we havediscussed a segmentation and feature extraction adaptive algorithm for structuredenvironments. In this study, several adaptive techniques proposed in literature forsegmentation of laser range data have been implemented and tested in different environmentsto compare the performances of them with the proposed technique. The experimental resultshave shown that the proposed method is superior to other adaptive techniques. Furtherdiscussion is continued to analyze the implementation issues of the artificial potential field approach in geometrical structured environments. The segmented features of the walls are used to generate the potential force for robot navigation. These segmented features arematched with the pre-observed features to extend or merge them together to generate a mapof the environment and this map is used in potential force generation process. Combining thesegmentation and representation of geometrical obstacles for artificial potential fieldgeneration in robot path planning, simulation experiments were done and performances arecompared for the traditional and the proposed approach.Based on the simulation results from various case studies, we have concluded that theproposed artificial potential field method for mobile robot path planning is able to solve the dead-lock problems that are with traditional method. The segmentation and feature extraction algorithm proposed in this thesis has shown better performances than the existing methods by experimental results. Geometrical representation of the structured environment is used to implement the artificial potential field based path planning on the robot and implementation barriers are discussed.
展开▼
