Optimizing ORB-SLAM For Varied Weather Conditions Using Genetic Algorithm

Abstract

Abstract In the rapidly evolving domains of self-driving cars, the resilience of Simultaneous Localization and Mapping (SLAM) algorithms to varying environmental conditions remains a critical challenge. This paper leverages the CARLA simulator to create comprehensive datasets that encompass an array of weather scenarios, ranging from clear sky to complex combinations of fog and rain, during both daytime and nighttime. The primary objective of this study is to optimize the performance of ORB-SLAM2 under these harsh conditions, improving resilience and robustness against different weather conditions. The evaluation is conducted using the Root Mean Square Error (RMSE) as the key metric. Genetic Algorithm (GA) is developed to optimize the parameters of ORB-SLAM. The GA aims to reduce the RMSE for each unique weather situation. The results show a significant improvement in ORB-SLAM’s performance and resilience, contributing to its potential applications in the broader landscape of autonomous systems and intelligent mobility networks.