Visual Simultaneous Localization and Mapping (VSLAM) is a prerequisite for robots to accomplish fully autonomous movement and exploration in unknown environments. At present, many impressive VSLAM systems have emerged, but most of them rely on the static world assumption, which limits their application in real dynamic scenarios. To improve the robustness and efficiency of the system in dynamic environments, this paper proposes a dynamic RGBD SLAM based on a combination of geometric and semantic information (DGS-SLAM). First, a dynamic object detection module based on the multinomial residual model is proposed, which executes the motion segmentation of the scene by combining the motion residual information of adjacent frames and the potential motion information of the semantic segmentation module. Second, a camera pose tracking strategy using feature point classification results is designed to achieve robust system tracking. Finally, according to the results of dynamic segmentation and camera tracking, a semantic segmentation module based on a semantic frame selection strategy is designed for extracting potential moving targets in the scene. Extensive evaluation in public TUM and Bonn datasets demonstrates that DGS-SLAM has higher robustness and speed than state-of-the-art dynamic RGB-D SLAM systems in dynamic scenes.

The ability of an autonomous Unmanned Aerial Vehicle (UAV) in an unknown environment is a prerequisite for its execution of complex tasks and is the main research direction in related fields. The autonomous navigation of UAVs in unknown environments requires solving the problem of autonomous exploration of the surrounding environment and path planning, which determines whether the drones can complete mission-based flights safely and efficiently. Existing UAV autonomous flight systems hardly perform well in terms of efficient exploration and flight trajectory quality. This paper establishes an integrated solution for autonomous exploration and path planning. In terms of autonomous exploration, frontier-based and sampling-based exploration strategies are integrated to achieve fast and effective exploration performance. In the study of path planning in complex environments, an advanced Rapidly Exploring Random Tree (RRT) algorithm combining the adaptive weights and dynamic step size is proposed, which effectively solves the problem of balancing flight time and trajectory quality. Then, this paper uses the Hermite difference polynomial to optimization the trajectory generated by the RRT algorithm. We named proposed UAV autonomous flight system as Frontier and Sampling-based Exploration and Advanced RRT Planner system (FSEPlanner). Simulation performs in both apartment and maze environment, and results show that the proposed FSEPlanner algorithm achieves greatly improved time consumption and path distances, and the smoothed path is more in line with the actual flight needs of a UAV.

Path planning is one of the key parts of unmanned aerial vehicle (UAV) fast autonomous flight in cluttered environments. However, it remains a challenge to efficiently generate a high-quality trajectory for flight tasks with a high success rate. In this paper, a robust planning framework is proposed, which can stably support autonomous flight tasks in complex unknown environments with limited onboard computing resources. Firstly, we propose the directed frontier point information structure (DFP), which can roughly capture the frontier information of the explored environment. The planning direction of a local planner can be evaluated and rectified efficiently based on the DFP to avoid falling into traps with limited cost. Secondly, an adaptive fusion replanning method is designed to generate a high-quality trajectory efficiently by incorporating two optimization methods with different characteristics, which can both take advantage of different optimization methods while avoiding disadvantages as much as possible, but also adjust the focus of the optimization according to the actual situation to improve the success rate of the planning method. Finally, sufficient comparison and evaluation experiments in simulation environments are presented. Experimental results show the proposed method has better performance, especially in terms of adaptability and robustness, compared to typical and state-of-the-art methods in unknown complex scenarios. Moreover, the proposed system is integrated into a fully autonomous quadrotor, and the effectiveness of the proposed method is further evaluated by using the quadrotor in real-world environments.