Path Planning of a Robot with Uncertain Observations

Abstract

The purpose of the work is to assess the performance and further improve a solution to the problem of autonomous robot optimal path planning under uncertainty. The path finding happens on a 2D plane modeled by an overlaid lattice. The idea in the solution is to combine deterministic and stochastic approaches. First assuming complete knowledge of the environment, the deterministic path planning problem is solved resulting in an optimal path; after that knowing that there may also be some unmapped static or slowly and randomly moving obstacles present in the environment; the online stochastic solution uses dynamic programming method to solve the path finding with obstacle avoidance problem. The proposed solution was rigorously put to test with different parameters and under various configurations to evaluate its performance and identify its weaknesses. The results of conducted experiments revealed notable achievements along with excellent opportunities for improvements. Hence, attempts were made to seize those opportunities and enhance the performance of the solution. The outcomes of those efforts were 35 % increase in the success rate and reduction in the time required for the solution to reach its goal by over 97 %.