Lantern-Explorer: A Collision-Avoidance Autonomous Exploration Drone System Based on Laser SLAM with Optimized Hardware and Software

Abstract. The inherent high flexibility of drone platforms has positioned them as a powerful tool when combined with LiDAR technology for acquiring three-dimensional data in confined spaces. However, due to limitations in onboard resources, energy, and flight stability, improving autonomous exploration efficiency and mapping accuracy has remained a challenge. To address this, we propose Lantern-Explorer, an autonomous exploration drone system optimized for both hardware and software based on LiDAR SLAM, to balance exploration efficiency and mapping accuracy in complex environments. The hardware design includes a compact, highly maneuverable, and stable coaxial dual-rotor octocopter platform with passive collision avoidance capabilities. A custom-developed flight controller supports high-bandwidth IMU data feedback to enhance the precision of the tightly-coupled LiDAR-inertial mapping module. On the software side, we designed an adaptive LiDAR odometry accuracy controller to achieve precise flight attitude control, ensuring high-speed flight while maintaining stability. Additionally, we proposed the improved omnidirectional LiDAR perception algorithm, FUEL-360, for autonomous exploration. This algorithm, based on the LiDAR FOV model, optimizes the strategy for detecting unknown frontiers, improving the efficiency of boundary extraction and viewpoint generation. By employing a viewpoint classification strategy based on a dual-nested Traveling Salesman Problem, it reduces redundant backtracking during exploration, ensuring the rationality of local and global path planning and thereby enhancing overall exploration efficiency. To verify the effectiveness of the optimized hardware and software design, extensive experiments were conducted in complex environments such as forests, tunnels, and underground parking lots. Compared with existing platforms and methods, Lantern-Explorer demonstrated significant advantages in both exploration efficiency and mapping accuracy. Experimental results indicate that the system has substantial engineering potential in real-world applications, providing a comprehensive and innovative solution for autonomous drone exploration in complex environments. The relevant software and hardware resources will be open-sourced at https://github.com/R7AY/Dream-Lantern to promote further research in the field.