Evolutionary Optimization for Safe Navigation of an Autonomous Robot in Cluttered Dynamic Unknown Environments
Year: 2018
Type: article
Abstract: We present a path planning approach based on probabilistic methods for a robot to navigate in a cluttered, dynamic, unknown environment. There are dynamic obstacles moving around and static obstacles located in the map. The robot does not have any prior information about them but should be able to navigate through the map beginning from a known starting point and safely ending at a known target point. The only information the robot has is the location of the starting point and the target point and it uses sensory information to collect information about its surroundings. Our method is compared to the D* Lite algorithm and results are presented. In the last section, the parameters of the robot are optimized using biogeography-based optimization (BBO). This is an efficient multivariable optimizer and it is shown that the results of optimization achieve significant improvement in robot navigation performance. In this thesis, we show that using evolutionary optimization methods like BBO can reduce the risk of collision and the navigation time by about 25% each. The resulting risk of collision indicates safe navigation by the robot which leads to the conclusion that this is a feasible method for real-world robots.
Author: Arash Roshanineshat
Institution: Cleveland State University
Language: English
Subfield: Control and Systems Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
Funder: Smithsonian Institution
OpenAlex ID: W2895716509