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 pixelwise 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.INDEX TERMS Simultaneous localization and mapping, dynamic environment, semantic, visual-inertial system.
Background including a long-period fast illumination variation is commonly assumed to be foreground by mistake. To solve this problem, proposed is a semantic-based hierarchical Gaussian mixture model integrated with an illumination detection approach. First, autocorrelation-based features for broad identification of background lighting changes and foreground in short-term sequences are presented. Then, the hierarchical Gaussians representing different background illumination variations are maintained. The effectiveness of the proposed method is demonstrated using experiments on pedestrian detection in fast lighting change.Introduction: Background subtraction is a commonly used approach to foreground detection. However, fast illumination variation of background presents a challenge for accurate detection results of foreground. Ordinarily, fast lighting change is considered to appear in regions rather than at pixels. Therefore, many approaches focus on the integration of regional information into a pixel-wise background model. The Gaussian mixture model (GMM) [1] is the most representative background model incorporated with other methods [2-4], so we are mainly concerned with the GMM in this Letter. Tian et al. brought a regional texture measure into the GMM (TGMM) and claimed the solution to accurate foreground detection in fast illumination variation [2]. A hierarchical GMM (HGMM) of different scales was devised to discriminate fast illumination change from foreground [3]. Meanwhile, a memorising GMM (MGMM) was proposed to tackle fast lighting change at pixels in a scene [4]. In the MGMM, Gaussians representing background were labelled and recorded in long-term memory. Actually, fast lighting change results from temporal intensity variations at pixels rather than those in regions. Thus, it is advisable to memorise the former background states deriving from fast lighting change. However, it is unadvisable to label the similar background states excessively without any intended discrimination between different illumination variations.In this Letter, we propose a semantic-based hierarchical GMM (S-HGMM) for long-term memory of periodical fast lighting change. A criterion for illumination detection is first made to identify different lighting changes in short-term sequences. Then, we design a hierarchical GMM including a sequence of short-term Gaussians and a single long-term one tagged semantically with a label of fast lighting change according to the illumination detection results.Background: The GMM is a pixel-wise mixture of K Gaussian distributions representing the recent states. The probability of intensity X t is P(X t ) = K k=1 v k,t × h(X t , m k,t , S k,t ), where h obeys the Gaussian distribution. m k,t , S k,t and v k,t represent the current mean, the covariance matrix and the weight estimation of Gaussian k, respectively. The covariance matrix S k,t is referred to as S k,t = s k,t × I, assuming the red, green and blue pixel values to be independent. A new pixel value X t + 1 is checked against each Gaussi...
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.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.