Review and Frontier Exploration of Active SLAM

This work provides a comprehensive review of active Simultaneous Localization and Mapping (SLAM), a pivotal technology for mobile robots navigating unknown environments. Active SLAM uniquely integrates active perception with decisionmaking to enhance exploration efficiency and localization accuracy, enabling applications in autonomous driving, disaster response, and beyond. The study formulates the active SLAM problem using the Partially Observable Markov Decision Process (POMDP) and Markov Decision Process (MDP), establishing a robust mathematical framework for optimized decision-making. It offers an in-depth analysis of the three core submodules of modular active SLAM approaches-goal position identification, cost-to-goal calculation, and optimal action execution-detailing their principles and diverse methodologies. Furthermore, the review highlights cutting-edge advancements, including deep reinforcement learning and continuous-space optimization, while addressing challenges in semantic understanding specific to active SLAM. Despite significant progress, issues such as environmental adaptability and robust data association persist. This work underscores the need for future research to leverage emerging technologies to enhance active SLAM's performance and reliability in complex, dynamic environments, positioning it as a critical contribution to the broader SLAM domain.