A path planning and obstacle avoidance hybrid system using a connectionist network
Schuster, Christopher Emmet
Cheatham, John B., Jr.
Master of Science
Automated path planning and obstacle avoidance has been the subject of intensive research in recent times. Most efforts in the field of semiautonomous mobile-robotic navigation involve using Artificial Intelligence search algorithms on a structured environment to achieve either good or optimal paths. Other approaches, such as incorporating Artificial Neural Networks, have also been explored. By implementing a hybrid system using the parallel-processing features of connectionist networks and simple localized search techniques, good paths can be generated using only low-level environmental sensory data. This system can negotiate structured two- and three-dimensional grid environments, from a start position to a goal, while avoiding all obstacles. Major advantages of this method are that solution paths are good in a global sense and path planning can be accomplished in real time if the system is implemented in customized parallel-processing hardware. This system has been proven effective in solving two- and three-dimensional maze-type environments.
System science; Mechanical engineering; Electronics; Electrical engineering; Artificial intelligence