Quantitative Metrics To Estimate The Environment Complexity For Robot's Navigation Performance

For practical applications,the notion of environmental complexity plays a vital role in the design and performance of robots.The hardware and software intricacy of a robot for a specific application or task is subject to the robot's awareness about the layout and organization of its surroundings.This study investigated different approaches to assess the complexity of the environment quantitatively in terms of mobile robots perspective.An evaluation framework is proposed to assess the environmental complexity in a given environment.The framework includes the setup of the environment,selection of the metrics,optimization of the metrics,and the evaluation of robot navigation performance in terms of the environmental complexity measures.To account for the varying levels of complexity in a given environment,we explored two metrics: Entropy and Mean Shortest Path.The entropy and mean shortest path measure is calculated for the environments with varying levels of complexity.The results of the simulated experiments are compared and analyzed.The Depth First Search algorithm-based navigation task is implemented to correlate the two measures and navigation performance of a robot.Based on experiments,we can say that these two metrics give a very similar reflection of complexity and are positively correlated.However,the Mean Shortest Path measure is more accurate than the Entropy measure in terms of assessing navigation-based task performance.Besides,the metrics identified how sub-regions in the whole map contrast with each other in terms of complexity.The results indicated that the examined measures could be employed to estimate the global complexity of different environments.Moreover,they could be representative of the capabilities of mobile robots navigation performance in a given environment.