Improve Tracking Performance Research Articles

Global Multi-Scale Optimization and Prediction Head Attentional Siamese Network for Aerial Tracking

Siamese-based trackers have been widely used in object tracking. However, aerial remote tracking suffers from various challenges such as scale variation, viewpoint change, background clutter and occlusion, while most existing Siamese trackers are limited to single-scale and local features, making it difficult to achieve accurate aerial tracking. We propose the global multi-scale optimization and prediction head attentional Siamese network to solve this problem and improve aerial tracking performance. Firstly, a transformer-based multi-scale and global feature encoder (TMGFE) is proposed to obtain global multi-scale optimization of features. Then, the prediction head attentional module (PHAM) is proposed to add context information to the prediction head by adaptively adjusting the spatial position and channel contribution of the response map. Benefiting from these two components, the proposed tracker solves these challenges of aerial remote sensing tracking to some extent and improves tracking performance. Additionally, we conduct ablation experiments on aerial tracking benchmarks, including UAV123, UAV20L, UAV123@10fps and DTB70, to verify the effectiveness of the proposed network. The comparisons of our tracker with several state-of-the-art (SOTA) trackers are also conducted on four benchmarks to verify its superior performance. It runs at 40.8 fps on the GPU RTX3060ti.

Adaptive Robust Terminal Sliding Mode Control with Integral Backstepping Synthesized Method for Autonomous Ground Vehicle Control

Autonomous ground vehicles (AGVs) operating in complex environments face the challenge of accurately following desired paths while accounting for uncertainties, external disturbances, and initial conditions, necessitating robust and adaptive control strategies. This paper addresses the critical path-tracking task in AGVs through a novel control framework for multilevel speed AGVs, considering both structured and unstructured uncertainties. The control system introduced in this study utilizes a nonlinear adaptive approach by integrating integral backstepping with terminal sliding mode control (IBTSMC). By incorporating integral action, IBTSMC continuously adjusts the control input to minimize tracking errors, improving tracking performance. The hybridization of the terminal sliding mode method enables finite time convergence, robustness, and a chatter-free response with reduced sensitivity to initial conditions. Furthermore, adaptive control compensators are developed to ensure robustness against unknown but bounded external disturbances. The Lyapunov stability theorem is employed to guarantee the global asymptotic stability of the closed-loop system and the convergence of tracking errors to the origin within finite time. To validate the effectiveness of the proposed control scheme, high-fidelity cosimulations are conducted using CarSim and MATLAB. Comparative analysis is performed with other methods reported in the literature. The results confirm that the proposed controller demonstrates competitive effectiveness in path-tracking tasks and exhibits strong efficiency under various road conditions, parametric uncertainties, and unknown disturbances.

Learning dynamic spatial-temporal regularized correlation filter tracking with response deviation suppression via multi-feature fusion

Visual object tracking (VOT) for intelligent video surveillance has attracted great attention in the current research community, thanks to advances in computer vision and camera technology. Meanwhile, discriminative correlation filter (DCF) trackers garnered significant interest owing to their high accuracy and low computing cost. Many researchers have introduced spatial and temporal regularization into the DCF framework to achieve a more robust appearance model and further improve tracking performance. However, these algorithms typically set fixed spatial and temporal regularization parameters, which limit flexibility and adaptability under cluttered and challenging scenarios. To overcome these problems, in this work, we propose a new dynamic spatial–temporal regularization for the DCF tracking model that emphasizes the filter to concentrate on more reliable regions during the training stage. Furthermore, we present a response deviation-suppressed regularization term for responses to encourage temporal consistency and avoid model degradation by suppressing relative response changes between two consecutive frames. Moreover, we introduce a multi-memory tracking framework to exploit various features and each memory contributes to tracking the target across all frames. Significant experiments on the OTB-2013, OTB-2015, TC-128, UAV-123, UAVDT, and DTB-70 datasets have revealed that the performance thereof outperformed many state-of-the-art trackers based on DCF and deep-based frameworks in terms of tracking accuracy and tracking success rate.

Iterative learning control for continuous-time multi-agent differential inclusion systems with full learnability

The learnability of systems plays a crucial role in the feasibility of control objective. This study explores the learnability and iterative learning-based consensus control for continuous-time multi-agent differential inclusion systems. Unlike discrete-time systems, analysis for the relationship between output space and realizable output space that used to explore learnability is more tough since involving continuous-time system output. Using measure theory, the systems exhibit full learnability if and only if the input–output coupling matrix is of full-row rank. Considering the possibility of controller power dissipation and demand for improving tracking performance, iterative learning control with state feedback and an efficiency factor is proposed. The consensus performance for the proposed control scheme is strictly explored. Simulation on unmanned vehicles in freeway demonstrates the validity of the theoretical results.

Observer-Based Finite-Time Fault-Tolerant Control for Nonstrict-Feedback Nonlinear Systems With Multiple Uncertainties

In the fault-tolerant control (FTC) tasks of nonstrict-feedback nonlinear systems, unmeasurable states, disturbance, and actuator faults are recognized as the main factors that obstacle the effective controller design and, thus, the tracking performance improvement. To tackle these obstructions, a fuzzy observer is introduced to address the difficulties of the unmeasurable states involving nonstrict-feedback nonlinear systems by benefiting from the approximation property of fuzzy logic systems. Owing to the newly employed damping term in the intermediate control law being utilized to compensate for the possibly unlimited number of faults, the proposed FTC strategy is able to deal with actuator faults properly without imposing tighter requirements on the fault mechanism. To reach fast transient performance, stability of finite time is reached by exploiting the backstepping method. The investigated strategy ensures that all the responses of the systems are semiglobal practical finite-time stable. Meanwhile, the tracking error converges to a small neighborhood of the origin within finite time. In addition, to demonstrate its effectiveness, the provided approach is applied to the position tracking of a robotic system, which shows anticipated control performances in spite of various uncertainties.

Robust Visual Object Tracking Based on Feature Channel Weighting and Game Theory

Although the discriminative correlation filter- (DCF)-based tracker improves tracking performance, some object representation issues can still be further optimized. On the one hand, the DCF tracker’s deep convolutional features contain many noisy channels, and assigning the same weights to multiple channels cannot distinguish the importance of different channels. On the other hand, a simple weighted fusion approach cannot fully utilize the benefits of different feature types. We propose a visual object tracking algorithm based on adaptive channel weighting and feature game fusion to solve these problems. In this study, an adaptive channel weighting strategy is designed to assign suitable weights to each channel based on the average energy ratio of the target and background regions in the feature channels and prune the channels with low weights to improve feature robustness and reduce computational complexity. Simultaneously, the game theory concept is introduced in the multifeature fusion. The handcrafted features are combined with shallow and deep convolutional features according to feature complementarity. Then, the two combined features are seen as two sides of the game, continuously gamed during the tracking process to generate a feature model with a higher representation capacity. Extensive experiments are conducted on four mainstream visual tracking benchmark datasets, including OTB2015, VOT2018, LaSOT, and UAV123. The experimental results show that the proposed algorithm performs outstandingly compared to the state-of-the-art trackers.

Enhancing GNSS receiver performance with software-defined vector carrier tracking for rocket launching

It is well-known that the use of Vector Tracking Loops (VTL) in GNSS receivers can result in improved tracking performance and sensitivity, faster acquisition, and improved interference robustness. However, despite the availability of software-defined radio (SDR) GNSS receivers in practical applications, there is currently no open-source implementation of Vector Tracking techniques able to operate in real-time. This work introduces a VTL algorithm applied to track the Doppler shift and its rate, using the open-source GNSS-SDR framework. This leads to a novel real-time SDR GNSS VTL receiver capable to work with different COTS front-ends. The proposed implementation was tested in a challenging non-conventional, high dynamics positioning scenario to obtain the position and velocity vectors of a rotation-stabilized guided rocket. The scenario was simulated with a GNSS RF signal generator, captured with a COTS front-end and processed by GNSS-SDR. Results show that the algorithm and its proposed implementation are able to provide full tracking, even under demanding accelerations. The positioning performance is measured using standard metrics, and we provide a fully automated reproducibility path. The source code is released under the General Public License v3.0, and will contribute to GNSS-SDR's open-source code base.

Active-Feedback 3D Single-Molecule Tracking Using a Fast-Responding Galvo Scanning Mirror.

Real-time three-dimensional single-particle tracking (RT-3D-SPT) allows continuous detection of individual freely diffusing objects with high spatiotemporal precision by applying closed-loop active feedback in an optical microscope. However, the current tracking speed in RT-3D-SPT is primarily limited by the response time of the control actuators, impeding long-term observation of fast diffusive objects such as single molecules. Here, we present an RT-3D-SPT system with improved tracking performance by replacing the XY piezoelectric stage with a galvo scanning mirror with an approximately 5 times faster response rate (∼5 kHz). Based on the previously developed 3D single-molecule active real-time tracking (3D-SMART), this new implementation with a fast-responding galvo mirror eliminates the mechanical movement of the sample and allows a more rapid response to particle motion. The improved tracking performance of the galvo mirror-based implementation is verified through simulation and proof-of-principle experiments. Fluorescent nanoparticles and ∼1 kB double-stranded DNA molecules were tracked via both the original piezoelectric stage and new galvo mirror implementations. With the new galvo-based implementation, notable increases in tracking duration, localization precision, and the degree to which the objects are locked to the center of the detection volume were observed. These results suggest that faster control response elements can expand RT-3D-SPT to a broader range of chemical and biological systems.

Immersion and invariance based adaptive robust control for attitude tracking of spacecraft with input saturation

This work investigates attitude control of spacecraft with uncertainties and input saturation. An immersion and invariance based adaptive robust control (IIARC) is proposed to improve tracking performance of spacecraft. The proposed approach consists of an immersion and invariance (I&I) estimator, a robust adaptive law, and a terminal sliding mode control (TSMC). The continuous fast TSMC is designed to stabilize the system and obtain fast convergence rate. To estimate the disturbances and uncertainties, I&I methodology is introduced into the proposed approach. The robust adaptive law is applied such that the prior information of disturbances is not required. IIARC with saturation compensation (IIARCSC) is presented to reduce the effect of input saturation on the attitude tracking of spacecraft. It is proved that the developed approach could guarantee the finite-time convergence (FTC). Extensive comparative simulations demonstrate the effectiveness of the presented scheme.

Adaptive Neural Trajectory Tracking Control for n-DOF Robotic Manipulators With State Constraints

This paper proposes an adaptive neural trajectory tracking control scheme for n-DOF robotic manipulators subjected to parameter variations, unknown functions, and time-varying external disturbances. First, the computed torque control (CTC) method is designed to reduce the system's nonlinearity. Second, radial basis function neural networks (RBFNNs) are constructed to approximate the uncertainties due to parameter variations and unknown functions. It's also important to note that the RBFNN's centers and widths are defined by state constraints. As a result of the nonlinear disturbance observer (NDO), the RBFNNs' approximation errors and disturbances are estimated to further improve tracking performance. The barrier Lyapunov function (BLF) ensures the closed-loop system's stability, guaranteeing tracking performance while preventing state constraint violation. Furthermore, sensitivity analysis provides a ranking of the importance of design parameters in influencing dynamic responses. Finally, simulations on a 7-DOF robotic manipulator are performed to validate the effectiveness of the proposed method.

Disturbance observer based actor-critic learning control for uncertain nonlinear systems

This paper investigates the disturbance observer based actor-critic learning control for a class of uncertain nonlinear systems in the presence of unmodeled dynamics and time-varying disturbances. The proposed control algorithm integrates a filter-based design method with actor-critic learning architecture and disturbance observer to circumvent the unmodeled dynamic and the time-varying disturbance. To be specific, the actor network is employed to estimate the unknown system dynamic, the critic network is developed to evaluate the control performance, and the disturbance observer is leveraged to provide efficient estimation of the compounded disturbance which includes the time-varying disturbance and the actor-critic network approximation error. Consequently, high-gain feedback is avoided and the improved tracking performance can be expected. Moreover, a composite weight adaptation law for actor network is constructed by utilizing two types of signals, the cost function and the modeling error. Eventually, theoretical analysis demonstrates that the developed controller can guarantee bounded stability. Extensive simulations and experiments on a robot manipulator are implemented to validate the performance of the resulted control strategy.

Vision-Based Hand Detection and Tracking Using Fusion of Kernelized Correlation Filter and Single-Shot Detection

Hand detection and tracking are key components in many computer vision applications, including hand pose estimation and gesture recognition for human–computer interaction systems, virtual reality, and augmented reality. Despite their importance, reliable hand detection in cluttered scenes remains a challenge. This study explores the use of deep learning techniques for fast and robust hand detection and tracking. A novel algorithm is proposed by combining the Kernelized Correlation Filter (KCF) tracker with the Single-Shot Detection (SSD) method. This integration enables the detection and tracking of hands in challenging environments, such as cluttered backgrounds and occlusions. The SSD algorithm helps reinitialize the KCF tracker when it fails or encounters drift issues due to sudden changes in hand gestures or fast movements. Testing in challenging scenes showed that the proposed tracker achieved a tracking rate of over 90% and a speed of 17 frames per second (FPS). Comparison with the KCF tracker on 17 video sequences revealed an average improvement of 13.31% in tracking detection rate (TRDR) and 27.04% in object detection error (OTE). Additional comparison with MediaPipe hand tracker on 10 hand gesture videos taken from the Intelligent Biometric Group Hand Tracking (IBGHT) dataset showed that the proposed method outperformed the MediaPipe hand tracker in terms of overall TRDR and tracking speed. The results demonstrate the promising potential of the proposed method for long-sequence tracking stability, reducing drift issues, and improving tracking performance during occlusions.

Incremental Nonlinear Dynamic Inversion Attitude Control for Helicopter with Actuator Delay and Saturation

In this paper, an incremental nonlinear dynamic inversion (INDI) control scheme is proposed for the attitude tracking of a helicopter with model uncertainties, and actuator delay and saturation constraints. A finite integral compensation based on model reduction is used to compensate the actuator delay, and the proposed scheme can guarantee the semi-globally uniformly ultimately bounded tracking. The overall attitude controller is separated into a rate, an attitude, and a collective pitch controller. The rate and collective pitch controllers combine the proposed method and INDI to enhance the robustness to actuator delay and model uncertainties. Considering the dynamic of physical actuators, pseudo-control hedging (PCH) is introduced both in the rate and attitude controller to improve tracking performance. By using the proposed controller, the helicopter shows good dynamics under the multiple restrictions of the actuators.

Online learning of MPC for autonomous racing

A Learning-based Model Predictive Control (LMPC) algorithm is proposed for a Formula Student (FS) autonomous vehicle. The online learning algorithm has two distinct roles: to improve the dynamic model accuracy of the vehicle used in the MPC, while performing online tuning of the model predictive controller parameters. The developed controller is shown to reduce the total lap time through an iterative learning process as the vehicle progresses on track. To capture the full complexity of the nonlinear higher order dynamics, an Artificial Neural Network (ANN) complements the vehicle’s nominal model. The ANN is trained using an online supervised learning scheme based on past model prediction errors. Additionally, a Genetic Algorithm (GA) is used to iteratively find the optimal set of controller parameters that maximizes a reward function. Several simulation tests performed on real examples of competition tracks demonstrate the effectiveness of the approach. Moreover, it is shown that the combination of both online learning methods is able to significantly improve tracking performance of the FS vehicle, eventually reducing the total lap time by over 16%.

Harmonic and synchronous compound control of a dual-valve controlled single-rod electro-hydraulic actuator

This article presents a novel harmonic and synchronous compound control scheme based on the dual-valve parallel structure for controlling a single-rod electro-hydraulic actuator with high-precision control and large-flow requirements. Harmonic and synchronous compound control consists of an adaptive robust control layer and a control law allocation layer. In adaptive robust control layer, an adaptive robust control algorithm is designed to deal with the system nonlinearities and uncertainties via backstepping method. Meanwhile, dynamic surface control technique and “lumped flow” are introduced to solve the differential explosion problem and the multiple input-single output control problem. In control law allocation layer, the harmonic and synchronous compound strategy allocates the control law to two valves reasonably based on system flow and valve frequency characteristics. Furthermore, since the effect of the proportional valve dead-zone can be compensated effectively by combining the estimation of the dead-zone flow by adaptive term, the estimation error compensation by robust term, and the allocation method of control law allocation layer, improved tracking performance can be achieved. Finally, both simulations and experiments are demonstrated to illustrate the advantages of the proposed control scheme.

Introducing convex BiLSTM-based controller applied to 3-PSP spatial parallel robot manipulator

The idea of modeling first and then finding a control approach might have made sense when systems were changing on a vast scale, but not in today’s world where we are in danger of being run over on a highway called the information highway. The feedback phenomenon, which involves a system’s future being influenced by its past, presents a rational conflict in this context. However, we can move beyond these challenges by leveraging the system’s inherent ability to randomly repeat and re-create parts, allowing new features to emerge and the best repetition to become the standard. To address the mentioned challenges, this paper proposes a new structure based on bilateral long short-term memory (BiLSTM) to provide supervising feedback on the nonlinear components comprising friction, Coriolis, and Centrifugal forces in addition to gravity for adapting uncertain variations in the robotic system dynamics. A key characteristic of the proposed BiLSTM-based controller is the incorporation of memory. A new formulation is developed for determining when to recover old memories and to what extent, ensuring the controller remains both adaptive and efficient. Furthermore, given that for system identification, speed is vital; the proposed approach has dealt with the speed category by introducing a convex combination of BiLSTM layers. The stability of the whole structure has also been analyzed, and the updated learning rules have been derived. The simulation results show improved tracking performance and learning time compared to alternative approaches.

Adaptive Neural Tracking Control for Manipulators With Prescribed Performance Under Input Saturation

In this article, an improved adaptive neural network (NN) nonsingular terminal sliding mode control (NTSMC) scheme is proposed for prescribed-performance trajectory tracking of manipulators with unmodeled dynamics and input saturation. In order to reduce the adverse effect of input saturation due to the conflict between excessive control force and limited motor torque, an auxiliary system is constructed. With the help of prescribed performance functions, we develop an improved NN-based NTSMC strategy to achieve tunable prescribed tracking errors under limited control, where it does not need prior precise knowledge of uncertainties. Theoretically, the uniform ultimate boundedness of the closed-loop system is proved by using the Lyapunov function. Finally, extensive comparative experiments are carried out on a ROKAE platform of a multidegree-of-freedom manipulator, and the improved tracking performance of the proposed scheme is verified.

SiamSampler: Video-Guided Sampling for Siamese Visual Tracking

We present SiamSampler, the first to our knowledge investigating video sampling in visual object tracking. We observe that the random sampling applied in Siamese-based trackers cannot focus on important data or ensure data diversity, hindering the effective training of networks. This paper proposes the Video-Guided Sampling Strategy to solve the problems in random sampling from both inter and intra-video levels. At the inter-video level, we propose Modified Gaussian Sampling Strategy (MGSS) to automatically assign higher sampling probabilities to longer and more difficult videos and reduce the sampling probabilities of shorter and easier videos. At the intra-video level, the Farthest Image Pair Sampling Strategy (FPSS) is proposed to increase the diversity of training data. Extensive experiments on general benchmarks demonstrate the effectiveness of our method. Compared with the baseline model, our method improves tracking performance on five datasets, without affecting the testing speed.

Robust Adaptive Fuzzy Backstepping Control for 2-DOF Laboratory Helicopter System with Improved Tracking Performance

Robust Adaptive Fuzzy Backstepping Control for 2-DOF Laboratory Helicopter System with Improved Tracking Performance

Joint target assignment and power allocation in the netted C-MIMO radar when tracking multi-targets in the presence of self-defense blanket jamming

The netted radar system (NRS) has been proved to possess unique advantages in anti-jamming and improving target tracking performance. Effective resource management can greatly ensure the combat capability of the NRS. In this paper, based on the netted collocated multiple input multiple output (C-MIMO) radar, an effective joint target assignment and power allocation (JTAPA) strategy for tracking multi-targets under self-defense blanket jamming is proposed. An architecture based on the distributed fusion is used in the radar network to estimate target state parameters. By deriving the predicted conditional Cramer-Rao lower bound (PC-CRLB) based on the obtained state estimation information, the objective function is formulated. To maximize the worst case tracking accuracy, the proposed JTAPA strategy implements an online target assignment and power allocation of all active nodes, subject to some resource constraints. Since the formulated JTAPA is non-convex, we propose an efficient two-step solution strategy. In terms of the simulation results, the proposed algorithm can effectively improve tracking performance in the worst case.