Investigation On Path Planning For Autonomous Mobile Robots In Partially Observable Environment

Law, Cheng Quan (2017) Investigation On Path Planning For Autonomous Mobile Robots In Partially Observable Environment. Technical Report. UTeM, Melaka. (Submitted)

Text Investigation On Path Planning For Autonomous Mobile Robots In Partially Observable Environment - Law Cheng Quan - 24 pages.pdf - Submitted Version Download (493kB)

Abstract

The autonomous mobile robot uses partially observable Markov decision processes (POMDP) model for the shortest and the best path planning to reach a destination in a partially structured environment. POMDP model is applied to improve computational efficiency of path planning problem. Sensing and information processing is important in autonomous mobile robots. Path planning in the real world is difficult because of partial observability and dynamic changes in the environment. Computational complexity increases when more variables are involved. The Perseus algorithm is investigated and the outcomes such as value function, reward and number of vectors are evaluated on different POMDP problems. Perseus algorithm improves belief point collection and selection to compute for better value functions. The algorithm randomly explores the belief space of an environment and collect a set of reachable belief points which will be fixed throughout the algorithm. Then, new value functions are computed to update the belief points. The algorithm repeats until a convergence criterion is met. Varying number of states and actions have significant effects on value function and number of vectors. While reward depends on the value of reward state and cost state.

Actions (login required)

View Item