Autonomous exploration in large-scale and complex environments is a challenging task. As the size of the environment increases, the significant overhead of exploration algorithms could overwhelm the computational capability of mobile platforms, prohibiting timely response to environmental changes. Meanwhile, the quality of exploration paths becomes increasingly important in larger scenes, as poorly selected paths greatly reduce efficiency. In this letter, a systematic framework is proposed to explore large-scale unknown environments. To enable high-frequency planning, a fast preprocessing of environmental information is presented, providing fundamental information to support high-frequency path planning. An path optimization formulation that comprehensively considers key factors about fast exploration is introduced. Further, an heuristic algorithm is devised to solve the NP-hard optimization problem, which empirically finds optimal solution in real time. Simulation results show the run time of our method is significantly shorter than existing ones. Our method completes exploration with the least time and shortest movement distance compared to current state-of-the-art methods.

Constructing a high-quality dense map in real-time is essential for robotics, AR/VR, and digital twins applications. As Neural Radiance Field (NeRF) greatly improves the mapping performance, in this paper, we propose a NeRF-based mapping method that enables higher-quality reconstruction and real-time capability even on edge computers. Specifically, we propose a novel hierarchical hybrid representation that leverages implicit multiresolution hash encoding aided by explicit octree SDF priors, describing the scene at different levels of detail. This representation allows for fast scene geometry initialization and makes scene geometry easier to learn. Besides, we present a coverage-maximizing keyframe selection strategy to address the forgetting issue and enhance mapping quality, particularly in marginal areas. To the best of our knowledge, our method is the first to achieve high-quality NeRF-based mapping on edge computers of handheld devices and quadrotors in real-time. Experiments demonstrate that our method outperforms existing NeRF-based mapping methods in geometry accuracy, texture realism, and time consumption. The code will be released at https://github.com/SYSU-STAR/H2-Mapping .

To improve the accuracy and efficiency of 3D LiDAR mapping, real-time cooperative SLAM has been considered to explore large and complex areas. To merge the individual maps from multiple robots, it is crucial to identify the common areas and obtain alternative matches between them. However, data transmission, especially in sparse networks with narrow bandwidth and limited range, is a challenging issue for the above problem. Since the distribution manner is suitable for limited communication, we proposed a common framework of 3D real-time distributed cooperative SLAM to fill the community gap. Assuming that each robot can communicate with others, the presented framework consists of four key modules: place recognition, relative pose estimation, distributed graph optimization, and communication. Meanwhile, we developed a complete real-time distributed cooperative SLAM system, called RDC-SLAM, by integrating state-of-the-art components into the framework. For computation and data transmission efficiency, descriptor-based registration is used instead of the conventional point cloud matching. An intensity-based descriptor is developed to perform the place recognition and obtain the alternative matches, while an eigenvalue-based segment descriptor is applied to further refine the relative pose estimations between these alternative matches. A distributed graph optimization method is utilized to obtain the maximum likelihood of multi-trajectory estimation. A communication protocol is also designed to associate data among robots that are easy to deploy and have low network requirements. The RDC-SLAM is validated by real-world experiments and exhibits superior performance concerning accuracy, computation efficiency, and data efficiency.

Compared with using only a single type of robot, the use of drones and ground vehicles to jointly explore unknown areas can bring efficiency improvements. However, due to the difficulty of ground-aerial loop detection and especially the lack of ground-air datasets in large outdoor scenes, there is no good solution for ground-aerial cooperative localization technology right now. Therefore, this letter introduces a dataset to facilitate the development of multimodal Simultaneous Localization And Mapping (SLAM) as well as ground-air heterogeneous Cooperative SLAM (C-SLAM) algorithms. This dataset was collected by a group of ground and aerial robots equipped with Light Detection and Ranging (LiDAR), cameras, and Global Navigation Satellite/Inertial Navigation Systems (GNSS/INS). All sensors were well-calibrated and triggered by a self-made synchronization module, and the synchronization accuracy can reach millisecond level. The high-precision ground truth is obtained by post-processing the navigation information collected by the GNSS/INS system. There are considerable overlaps in the acquisition routes on the ground and in the air, which facilitates C-SLAM algorithms and loop detection. Also, the special installation angles in the air, and scenes such as channel and roof bring new challenges to the SLAM algorithm. We also tested several well-known algorithms on our dataset.

It is challenging for a mobile robot to avoid moving obstacles in dynamic environments. Traditional velocity obstacle methods do not fully consider the obstacles moving with the speeds larger than the maximum speed of the robot. In this paper, a new obstacle avoidance method, named the maximum-speed aware velocity obstacle (MVO) algorithm, is proposed for a mobile robot to avoid one or multiple high-speed obstacles. The proposed algorithm expands the velocity obstacle region into two parts, where one of the parts foresees collisions beyond the time horizon to ensure the feasible solutions of the current and the next control step. In practical applications, the perception capability of the robot is generally limited, and a non-holonomic robot can’t move into any direction due to its kinematic constraints. In this paper, the limited sensing field of view and non-holonomic kinematic constraints of the mobile robot are incorporated into the proposed MVO method. Moreover, continuity, safety, and computational complexity of the MVO approach are analyzed and presented. Extensive simulations and physical experiments are conducted to verify the efficacy of the MVO method, where a quadrotor and a differential-drive robot are used to perform dynamic obstacle avoidance.

A heterogeneous multi-robot system consisting of Unmanned Ground Vehicles (UGVs) and Unmanned Aerial Vehicles (UAVs) have advantages over a single-robot system in efficiency and flexibility, enabling them to perform a larger range of tasks. To allow heterogeneous platforms to work together in GPS-denied scenarios, it is crucial to build a complete 3D map of the environment. In this letter, a novel method is presented to perform ground and aerial collaborative mapping leveraging visual and range data collected by cameras and 3D LiDAR sensors. In the proposed system, a visual-LiDAR ego-motion estimation module that considers point, line and planar constraints can provide robust odometry information. Thumbnail images representing obstacle outlines are generated and descriptors are extracted using a neural network to help perform data association between separate runs. Map segments and the robot poses are organized together and are updated during a pose graph optimization procedure. The proposed ground-aerial collaborative mapping approach is evaluated on both synthetic and real-world datasets comparing with other methods. Experiment results demonstrate that our method can achieve outstanding mapping results.