Autonomous Exploration of Unknown Environments Using Entrance Detection, Classification, and Multi-Target Entrance Organization

Exploration in unknown environments remains a significant challenge in robotics and artificial intelligence, particularly in navigating dynamic (always changing) and complex terrains such as indoor areas with multiple rooms and obstacles. Traditional static target algorithms struggle with dynamic changes and lack the capability to verify detected entrances or junctions as actual free spaces. Additionally, these methods process one target at a time, increasing the risk of missing critical exploration areas. The need for advanced algorithms that adapt to dynamic entrance, evaluate detected entrance, sort multiple entrances, and efficiently explore based on importance is essential for advancing both research and practical applications. To address these challenges, this paper introduces the Multiple Target Evaluation Node (M-TEN) framework, inspired by Frontier Exploration. Using LiDAR data, M-TEN detects possible entrances, verifies true entrances with 99% accuracy, and optimally generates a sequence for multiple entrances. Performance was tested in a 10m x 10m simulation environment with two map configurations. The results demonstrate that M-TEN enables mobile robots to explore unknown environments efficiently, achieving 94.5% mapping coverage within 3.8 minutes compared to manually controlled and basic frontier exploration approach. This framework enhances exploration accuracy, reduces time, and ensures reliable target evaluation and sorting.