Cooperative Visual-SLAM System for UAV-Based Target Tracking in GPS-Denied Environments: A Target-Centric Approach

Trujillo, Juan-Carlos; Munguía, Rodrigo; Urzua, Sarquis; Grau, Antoni

doi:10.3390/electronics9050813

Cited by 7 publications

(6 citation statements)

References 59 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Various attempts have been made to solve the object tracking problem using traditional decision-making systems, such as the Visual Simultaneous Localization and Mapping (VSLAM) method [10]. However, the VSLAM method is limited in generalization because it is based on a pre-set model.…”

Section: Related Workmentioning

confidence: 99%

UAV Dynamic Object Tracking with Lightweight Deep Vision Reinforcement Learning

Nguyen

Thudumu

et al. 2023

Algorithms

View full text Add to dashboard Cite

Several approaches have applied Deep Reinforcement Learning (DRL) to Unmanned Aerial Vehicles (UAVs) to do autonomous object tracking. These methods, however, are resource intensive and require prior knowledge of the environment, making them difficult to use in real-world applications. In this paper, we propose a Lightweight Deep Vision Reinforcement Learning (LDVRL) framework for dynamic object tracking that uses the camera as the only input source. Our framework employs several techniques such as stacks of frames, segmentation maps from the simulation, and depth images to reduce the overall computational cost. We conducted the experiment with a non-sparse Deep Q-Network (DQN) (value-based) and a Deep Deterministic Policy Gradient (DDPG) (actor-critic) to test the adaptability of our framework with different methods and identify which DRL method is the most suitable for this task. In the end, a DQN is chosen for several reasons. Firstly, a DQN has fewer networks than a DDPG, hence reducing the computational resources on physical UAVs. Secondly, it is surprising that although a DQN is smaller in model size than a DDPG, it still performs better in this specific task. Finally, a DQN is very practical for this task due to the ability to operate in continuous state space. Using a high-fidelity simulation environment, our proposed approach is verified to be effective.

show abstract

Section: Related Workmentioning

confidence: 99%

UAV Dynamic Object Tracking with Lightweight Deep Vision Reinforcement Learning

Nguyen

Thudumu

et al. 2023

Algorithms

View full text Add to dashboard Cite

show abstract

“…The map matching technology uses external sensors of the platform to detect the surrounding environment information and compare it with a built-in map for its own matching and positioning, which has high time and space complexity and puts high requirements on the platform's own performance, making it difficult to meet the requirements of real-time scenarios. The projection tracking technology relies on real-time calculation of the platform's own displacement for projection positioning, and the main methods are Inertial Navigation System (INS) [8,9], Visual Odometry (VO) [10,11], and Simultaneous Localization and Mapping (SLAM) [12][13][14], but these methods are unable to overcome the effect of cumulative error, which makes it difficult to maintain high-accuracy positioning over long periods of Aerospace 2024, 11, 162 2 of 22 time. The cooperative positioning method based on auxiliary beacons [15,16] relies on auxiliary beacons with known high-precision coordinates, which will fail in the complex environment once the platform maneuvers or moves away from auxiliary beacons.…”

Section: Introductionmentioning

confidence: 99%

A Novel Sea Target Tracking Algorithm for Multiple Unmanned Aerial Vehicles Considering Attitude Error in Low-Precision Geodetic Coordinate Environments

Dai,

Lu,

2024

Aerospace

View full text Add to dashboard Cite

Geodetic coordinate information and attitude information of the observation platform are necessary for multi-UAV position alignment and target tracking. In a complex sea environment, the navigation equipment of a UAV is susceptible to interference. High-precision geodetic coordinate information and attitude information are difficult to obtain. Aiming to solve the above problems, a low-precision geodetic coordinate real-time systematic spatial registration algorithm based on multi-UAV observation and an improved robust fusion tracking algorithm of multi-UAV to sea targets considering attitude error are proposed. The spatial registration algorithm obtains the observation information of the same target based on the mutual observation information. Then, geodetic coordinate systematic error is accurately estimated by establishing the systematic error estimation measurement equation. The improved robust fusion tracking algorithm considers the influence of UAV attitude error in the observation. The simulation experiment and practical experiment show that the algorithm can not only estimate systematic error accurately but also improve tracking accuracy.

show abstract

“…Compared to the conventional wheeled odometry, VO has comprehensive advantages in terms of cost, accuracy, and reliability. It constitutes the foundation of the visual positioning systems such as simultaneous localization and mapping (SLAM) and structure from motion (SFM) [1][2][3][4][5][6].…”

Section: Introductionmentioning

confidence: 99%

“…Our main contributions are as follows. (1) We propose an R-CNN architecture to learn effective latent OF representation and further to model OF dynamics and sequential relations so that the motion estimation is constrained by the relations between sequential OF features, thereby alleviating errors. (2) We develop an encoder-decoder architecture to train the OF encoder separately in an unsupervised manner.…”

Section: Introductionmentioning

confidence: 99%

Ego-Motion Estimation Using Recurrent Convolutional Neural Networks through Optical Flow Learning

Zhao

Wei

2021

Electronics

View full text Add to dashboard Cite

Visual odometry (VO) refers to incremental estimation of the motion state of an agent (e.g., vehicle and robot) by using image information, and is a key component of modern localization and navigation systems. Addressing the monocular VO problem, this paper presents a novel end-to-end network for estimation of camera ego-motion. The network learns the latent subspace of optical flow (OF) and models sequential dynamics so that the motion estimation is constrained by the relations between sequential images. We compute the OF field of consecutive images and extract the latent OF representation in a self-encoding manner. A Recurrent Neural Network is then followed to examine the OF changes, i.e., to conduct sequential learning. The extracted sequential OF subspace is used to compute the regression of the 6-dimensional pose vector. We derive three models with different network structures and different training schemes: LS-CNN-VO, LS-AE-VO, and LS-RCNN-VO. Particularly, we separately train the encoder in an unsupervised manner. By this means, we avoid non-convergence during the training of the whole network and allow more generalized and effective feature representation. Substantial experiments have been conducted on KITTI and Malaga datasets, and the results demonstrate that our LS-RCNN-VO outperforms the existing learning-based VO approaches.

show abstract

Cooperative Visual-SLAM System for UAV-Based Target Tracking in GPS-Denied Environments: A Target-Centric Approach

Cited by 7 publications

References 59 publications

UAV Dynamic Object Tracking with Lightweight Deep Vision Reinforcement Learning

UAV Dynamic Object Tracking with Lightweight Deep Vision Reinforcement Learning

A Novel Sea Target Tracking Algorithm for Multiple Unmanned Aerial Vehicles Considering Attitude Error in Low-Precision Geodetic Coordinate Environments

Ego-Motion Estimation Using Recurrent Convolutional Neural Networks through Optical Flow Learning

Contact Info

Product

Resources

About