UAV swarm applications are critical for the future, and their mission-planning and decision-making capabilities have a direct impact on their performance. However, creating a dynamic and scalable assignment algorithm that can be applied to various groups and tasks is a significant challenge. To address this issue, we propose the Extensible Multi-Agent Deep Deterministic Policy Gradient (Ex-MADDPG) algorithm, which builds on the MADDPG framework. The Ex-MADDPG algorithm improves the robustness and scalability of the assignment algorithm by incorporating local communication, mean simulation observation, a synchronous parameter-training mechanism, and a scalable multiple-decision mechanism. Our approach has been validated for effectiveness and scalability through both simulation experiments in the Multi-Agent Particle Environment (MPE) and a real-world experiment. Overall, our results demonstrate that the Ex-MADDPG algorithm is effective in handling various groups and tasks and can scale well as the swarm size increases. Therefore, our algorithm holds great promise for mission planning and decision-making in UAV swarm applications.
Simultaneous localization and mapping(SLAM), focusing on addressing the joint estimation problem of self-localization and scene mapping, has been widely used in many applications such as mobile robot, drone, and augmented reality(AR). However, traditional state-of-the-art SLAM approaches are typically designed under the static-world assumption and prone to be degraded by moving objects when running in dynamic scenes. This paper presents a novel semantic visual-inertial SLAM system for dynamic environments that, building on VINS-Mono, performs real-time trajectory estimation by utilizing the pixel-wise results of semantic segmentation. We integrate the feature tracking and extraction framework into the front-end of the SLAM system, which could make full use of the time waiting for the completion of the semantic segmentation module, to effectively track the feature points on subsequent images from the camera. In this way, the system can track feature points stably even in high-speed movement. We also construct the dynamic feature detection module that combines the pixel-wise semantic segmentation results and the multi-view geometric constraints to exclude dynamic feature points. We evaluate our system in public datasets, including dynamic indoor scenes and outdoor scenes. Several experiments demonstrate that our system could achieve higher localization accuracy and robustness than state-of-the-art SLAM systems in challenging environments.
26 October 2023 in IEEE Transactions on Industrial Electronics
This article introduces visual-depth-inertial-wheel odometry (VDIWO), a robust approach for real-time localization of mobile robots in indoor and outdoor scenarios. Notably, VDIWO achieves accurate localization without relying on prior information. This approach integrates the RGB-D camera, inertial measurement unit, and odometer measurements in a tightly coupled optimization framework. First, we introduce the depth measurement model based on Gaussian mixed model to predict the depth uncertainty of feature points. Then, we propose a hybrid depth estimation method that utilizes both depth measurement fusion and multiview triangulation to estimate the depth of landmarks and simultaneously identify high-quality landmarks. Furthermore, we integrate visual reprojection with depth measurement constraints and odometer preintegration constraints into the tightly coupled optimization framework to further enhance pose estimation accuracy. We evaluate the performance of the VDIWO method using OpenLORIS datasets and real-world experiments. The results demonstrate the high accuracy and robustness of VDIWO for state estimation of mobile robots.
In challenging environments, feature-based visual SLAM encounters frequent failures in frame tracking, introducing unknown poses to robotic applications. This paper introduces an immediate approach for recovering untracked camera poses. Through the retrieval of key information from elapsed untracked frames, lost poses are efficiently restored with a short time consumption. Taking account of reconstructed poses and map points during local optimizing, a denser local map is constructed around ambiguous frames to enhance the further SLAM procedure. The proposed method is implemented in a SLAM system, and monocular experiments are conducted on datasets. The experimental results demonstrate that our method can reconstruct the untracked frames in nearly real time, effectively complementing missing segments of the trajectory. Concurrently, the accuracy and robustness for subsequent tracking are improved through the integration of recovered poses and map points.
The efficacy of visual Simultaneous Localization and Mapping (SLAM) diminishes in large-scale environments due to challenges in identifying distant landmarks, leading to a limited perception range and trajectory drift. This paper presents a practical method to enhance the accuracy of feature-based real-time visual SLAM for compact unmanned vehicles by constructing distant map points. By tracking consecutive image features across multiple frames, remote map points are generated with sufficient parallax angles, extending the mapping scope to the theoretical maximum range. Observations of these landmarks from preceding keyframes are supplemented accordingly, improving back-end optimization and, consequently, localization accuracy. The effectiveness of this approach is ensured by the introduction of the virtual map point, a proposed data structure that links relational features to an imaginary map point, thereby maintaining the constrained size of local optimization during triangulation. Based on the ORB-SLAM3 code, a SLAM system incorporating the proposed method is implemented and tested. Experimental results on drone and vehicle datasets demonstrate that the proposed method outperforms ORB-SLAM3 in both accuracy and perception range with negligible additional processing time, thus preserving real-time performance. Field tests using a UGV further validate the efficacy of the proposed method.