Multiple Vehicle Tracking Research Articles

Retraction Note: An intelligent multiple vehicle detection and tracking using modified vibe algorithm and deep learning algorithm

Retraction Note: An intelligent multiple vehicle detection and tracking using modified vibe algorithm and deep learning algorithm

Distributed avoidance-accommodating flexible performance approach for adaptive formation tracking of underactuated surface vehicles against moving obstacles

A novel strategy for formation tracking of multiple uncertain underactuated surface vehicles is introduced in this study, to maintain the prescribed formation performance in the presence of moving obstacles with unknown velocities. Unlike the existing methods for prescribed formation performance control, our key contribution is the development of distributed avoidance-accommodating flexible performance functions (FPFs). These functions prevent deviation from the prescribed performance bounds while navigating around moving obstacles, considering input saturation and model uncertainties. We design local adaptive formation controllers by using distributed posture errors and avoidance-accommodating FPFs. Auxiliary signals that facilitate adjustment of the posture errors and performance functions during obstacle avoidance are derived using a recursive design that ensures Lyapunov stability. Additionally, command filters for virtual control laws are redesigned as saturation-compensating filters to solve the input saturation problem. Consequently, the proposed controller ensures the prescribed distributed formation performance, even in scenarios involving moving obstacles and input saturation, a capability lacking in conventional prescribed performance-based formation controllers, which can become unstable in such scenarios. The theoretical findings of this study are validated through comparative simulations.

Distributed formation tracking control of multiple autonomous surface vehicles: A hierarchical event-triggered approach

This paper proposes a hierarchical event-triggered approach for distributed formation tracking of multiple autonomous surface vehicles (ASVs). The kinematic and dynamic loops are developed to generate the guidance law and the control efforts respectively. In the kinematic loop, a surrogate model is developed for each ASV, so that the kinematic states of the ASV can be predicted by its neighbors. Only when a predefined event-triggering condition is satisfied, will the ASV’s true states be broadcast for updating the corresponding surrogate model, which circumvents the continuous states exchanges between ASVs. Additionally, a distributed leader motion observer is constructed to estimate the virtual leader’s velocities to avoid acquiring global information. In the dynamic loop, a novel event-triggered extended state observer is developed, by which the velocity tracking errors and the model uncertainties are estimated online and transmitted to a dynamic-inversion based controller only when necessary. The communication loads in sensor-to-controller and controller-to-actuator channels are thus simultaneously reduced. The stability of the closed-loop system is rigorously proved by the Lyapunov theory and cascade system theory. The Zeno behavior is also well excluded in multiple channels. Simulation results show that the proposed scheme has satisfactory control performances and distinct features of low communication burdens and energy consumption.

Multitarget-Tracking Method Based on the Fusion of Millimeter-Wave Radar and LiDAR Sensor Information for Autonomous Vehicles.

Multitarget tracking based on multisensor fusion perception is one of the key technologies to realize the intelligent driving of automobiles and has become a research hotspot in the field of intelligent driving. However, most current autonomous-vehicle target-tracking methods based on the fusion of millimeter-wave radar and lidar information struggle to guarantee accuracy and reliability in the measured data, and cannot effectively solve the multitarget-tracking problem in complex scenes. In view of this, based on the distributed multisensor multitarget tracking (DMMT) system, this paper proposes a multitarget-tracking method for autonomous vehicles that comprehensively considers key technologies such as target tracking, sensor registration, track association, and data fusion based on millimeter-wave radar and lidar. First, a single-sensor multitarget-tracking method suitable for millimeter-wave radar and lidar is proposed to form the respective target tracks; second, the Kalman filter temporal registration method and the residual bias estimation spatial registration method are used to realize the temporal and spatial registration of millimeter-wave radar and lidar data; third, use the sequential m-best method based on the new target density to find the track the correlation of different sensors; and finally, the IF heterogeneous sensor fusion algorithm is used to optimally combine the track information provided by millimeter-wave radar and lidar, and finally form a stable and high-precision global track. In order to verify the proposed method, a multitarget-tracking simulation verification in a high-speed scene is carried out. The results show that the multitarget-tracking method proposed in this paper can realize the track tracking of multiple target vehicles in high-speed driving scenarios. Compared with a single-radar tracker, the position, velocity, size, and direction estimation errors of the track fusion tracker are reduced by 85.5%, 64.6%, 75.3%, and 9.5% respectively, and the average value of GOSPA indicators is reduced by 19.8%; more accurate target state information can be obtained than a single-radar tracker.

On novel distributed fixed-time formation tracking of multiple hypersonic flight vehicles with collision avoidance

This paper aims to achieve collision-free formation tracking for multiple hypersonic flight vehicles without introducing conventional approximators such as fuzzy logic systems and neural networks. By relying on the hierarchical control theory, a globally fixed-time distributed formation control protocol with guaranteed tracking behaviors is proposed. Firstly, for the purpose of deriving the virtual control commands in the three-dimensional translational dynamics, a fixed-time consensus control input is presented. Then, by designing a comprehensive performance-preserving feedback control law, the actual control commands are derived, so that the virtual control inputs can be tracked timely and accurately. Meanwhile, with resort to a novel exponential-type artificial potential function, a collision-free control input is proposed to avoid close proximity between adjacent hypersonic flight vehicles. Summarily, the outstanding advantage of the proposed protocol is that it simultaneously realize the collision-free formation tracking, fixed-time stability and performance-preserving control for multiple hypersonic flight vehicles with uncertain dynamics. Lyapunov stability criterion verifies the closed-loop stability, showing that the error signals can attenuate to a tiny region around the origin within a user-specified time. Finally, comparative numerical simulations are performed to further demonstrate the theoretical results.

Selection strategies and finite-time target tracking of multiple unmanned surface vehicles with mode uncertainty and disturbances

In this paper, the finite-time coordinated target tracking problem of multi-unmanned surface vehicles (USVs) with model uncertainty, underactuation and disturbances based on distributed coordinated control is studied. Firstly, from the point of view of practical application, the select tracking scenario is proposed to protect monitoring area. Three new initial selection strategies are proposed to shorten the total time for the whole tracking system. Secondly, based on sliding mode control, the surge force finite time tracking controller and the yaw moment finite time tracking controller are proposed. And the novel continuous function is introduced to eliminate the singularity. Finally, the Lyapunov stability criterion proves that the designed controllers can realize multi-USVs coordinated finite-time tracking. Simulation results show the effectiveness of the designed distributed coordinated controllers, and verify the correctness of the initial selection strategies.

Cooperative learning formation control of multiple autonomous underwater vehicles with prescribed performance based on position estimation

This paper proposes a cooperative learning formation control method with finite-time prescribed performance based on position estimation for parametric path tracking of multiple autonomous underwater vehicles (AUVs) with uncertainties and external disturbances. The parametric path used in the control law permits the velocity to be specified independently while tracking the path accurately. The localized radial basis function neural networks learn the uncertainties cooperatively while tracking the period path, and the knowledge gained from learning is utilized to construct an empirically based formation control law using experience to cope with similar uncertainties rather than repeatedly using adaptive methods, which reduces the computing burden. The position of the leader is assumed to be available only for the leader's neighboring AUV, and a novel finite-time distributed observer is presented for the followers to estimate the leader's position. Based on this, the control law is derived from the prescribed performance control method using a finite-time performance function rather than exponential decaying function to enable the tracking error converges in finite time, which accelerates the learning process. The simulation results confirm the validity of the presented control protocol.

Cooperative target search and tracking for multi-UAVs based on control barrier functions

In this paper, a multiple unmanned aerial vehicles (multi-UAVs) cooperative target search and tracking strategy with collision avoidance and flocking constraints are presented applying the control barrier functions (CBFs). First, the constraints of collision avoidance and flocking maintenance are designed based on CBF. Second, the information map and Lyapunov guidance vector field (LGVF) are used to design the corresponding nominal controllers in search mode and tracking mode, respectively. To minimize the difference between the actual controller and the nominal controller, the controller based on quadratic programming (QP) is proposed to achieve the target search and tracking on the premise of security. Finally, the simulation results verify the effectiveness of the proposed method.

Multiple vehicles tracking and detection using weight high order singular value decomposition dimensionality reduction and double classifiers

Controlling and managing city traffic is one of them. In order to use image processing to prevent accidents on the road, vehicle tracking and detection are essential. By following moving objects, surveillance video monitoring and human activity recording are carried out. By taking this into account, a useful technique for image processing that detects automobiles from the image is suggested. For numerous vehicle tracking and detection systems, the ECNN-SVM (Enhanced Convolution Neural Network with Support Vector Machine) has just been introduced. However, the larger dimensional data space and inaccurate edge recognition make this system’s performance difficult. The WHOSVD (Weight High Order Singular Value Decomposition) approach, which reduces the dimension and breaks up the positive and negative training picture samples, is established to improve training speed and visual vehicle recognition. To effectively identify the edges at corners, improved canny edge detection is used for edge detection. Mean Kernel Fuzzy C Means (MKFCM) clustering algorithm-based three-dimensional bounding box estimation is used to identify the vehicle items. By merging the feature value of samples with their class labels, the Speed Factor Based Cuckoo Search Algorithm (SFCSA) is introduced for feature selection. The WHOSVD algorithm was used as the input for the enhanced convolutional neural network (ECNN), which is introduced for low-dimensional space and is used for vehicle detection and tracking. Occlusion problems are resolved and target features are further identified using a machine learning classifier. For common algorithms like CNN+SVM, Support Vector Machine (SVM), and the proposed technique, experimentation is done in regards to the metrics of accuracy, f-measure, precision, and recall for performance evaluation.

Cooperative Target Tracking of Multiple Autonomous Surface Vehicles Under Switching Interaction Topologies

This paper is concerned with the cooperative target tracking of multiple autonomous surface vehicles (ASVs) under switching interaction topologies. For the target to be tracked, only its position can be measured/received by some of the ASVs, and its velocity is unavailable to all the ASVs. A distributed extended state observer taking into consideration switching topologies is designed to integrally estimate unknown target dynamics and neighboring ASVs' dynamics. Accordingly, a novel kinematic controller is designed, which takes full advantage of known information and avoids the approximation of some virtual control vectors. Moreover, a disturbance observer is presented to estimate unknown time-varying environmental disturbance. Furthermore, a distributed dynamic controller is designed to regulate the involved ASVs to cooperatively track the target. It enables each ASV to adjust its forces and moments according to the received information from its neighbors. The effectiveness of the derived results is demonstrated through cooperative target tracking performance analysis for a tracking system composed of five interacting ASVs.

Vision-Assisted 3-D Predictive Beamforming for Green UAV-to-Vehicle Communications

Unmanned aerial vehicles (UAVs) have demonstrated exceptional capabilities in various applications such as vehicle tracking, military monitoring and communication assistance. With their advantages in mobile communication and integrated multiple sensors, UAVs have great potential to obtain the integration gain through integrated sensing and communications (ISAC) designs. In this paper we focus on the beamforming algorithm for UAV-to-vehicle communications. To reduce the communication overhead of block transmission caused by the high mobility of the target vehicles and the UAV, we utilize the inherent vision functionality of the UAV platform and propose a vision-assisted beamforming framework. Specifically, to obtain the position of to-be-served vehicles, we employ YOLOv5, a deep learning model, for vehicle detection, followed by a Sage-Husa filter for multiple target vehicle tracking. Based on the predicted positions of the vehicles, we propose a lightweight beamforming algorithm to save communication overhead of beam tracking. To validate the effectivess of this work, we perform experiments on the UAVDT dataset for vehicle detection and tracking. The simulation results show that the proposed algorithm achieves significant performance gains on both the effective achievable rates and the energy efficiency.

Consensus Formation Tracking for Multiple AUV Systems Using Distributed Bioinspired Sliding Mode Control

Consensus formation tracking of multiple autonomous underwater vehicles (AUVs) subject to nonlinear and uncertain dynamics is a challenging problem in robotics. To tackle this challenge, a distributed bioinspired sliding mode controller is proposed in this paper. First, the conventional sliding mode controller (SMC) is presented, and the consensus problem is addressed on the basis of graph theory. Next, to tackle the high frequency chattering issue in SMC scheme and meanwhile improve the robustness to the noises, a bioinspired approach is introduced, in which a neural dynamic model is employed to replace the nonlinear sign or saturation function in the synthesis of conventional sliding mode controllers. Furthermore, the input-to-state stability of the resulting closed-loop system is proved in the presence of bounded lumped disturbance by the Lyapunov stability theory. Finally, simulation experiments are conducted to demonstrate the effectiveness of the proposed distributed formation control protocol.

Artificial Potential Field-Based Multi-UAV Formation Control and Target Tracking

To simultaneously achieve space formation flight and target tracking of multiple unmanned aerial vehicles (UAVs) and solve the rotation buffeting problem of the UAV, a robust formation control and target tracking algorithm is proposed. The artificial potential function consisting of formation control term and target tracking term is established, and its convergence is proved. The sliding mode control method with the saturation function is established, and a sufficient condition for sliding mode to occur is analysed. Finally, the numerical simulation is conducted for the proposed algorithm, and the simulation result is analysed. The results show that the proposed algorithm can quickly achieve the formation flight and target tracking of multi-UAVs and improve the tracking performance; meanwhile, it can effectively weaken the rotation buffeting and improve the robustness.

Quantized-communication-based neural network control for formation tracking of networked multiple unmanned surface vehicles without velocity information

This paper proposes a quantized communication-based output-feedback control strategy for formation tracking of networked unmanned surface vehicles (USVs) with uncertainty. Under limited network communication, it is assumed that each USV measures only the position and orientation information. In particular, this information is quantized and transmitted to USVs connected to a band-limited directed network. The primary contributions of this study are to derive distributed learning laws for neural networks using discontinuous signals and to analyze the stability of the neural network-based output-feedback control system designed in a quantized communication environment. A neural network-based local observer is developed to estimate the velocity information of each USV with model uncertainty and external disturbance. Then, a neural network-based output-feedback control design strategy using distributed and quantized postures is presented to accomplish the desired formation of networked USVs with uncertainty and underactuation. The distributed learning laws of neural networks are derived using neighbors’ quantized signals. The auxiliary signal and approach angle are employed to solve the underactuation and stability analysis problems. Despite the discontinuity of quantized signals, it is proven that all errors in the closed-loop system are bounded and can be made arbitrarily small. Finally, simulation results are given to verify the theoretical results of the proposed control system.

Self-Driving Car

Abstract: Every person in this world is concerned about being safe. Increasing safety and reducing road accidents, thereby saving lives are of great interest in the context of Advanced Driver Assistance Systems. Among the complex and challenging tasks of future road vehicles is road lane detection or road boundaries detection. In driving assistance systems, obstacle detection especially for moving object detection is a key component of collision avoidance. Many sensors can be used for obstacle detection and lane detection, such as laser, radar and vision sensors. The most frequently used principal approach to detect road boundaries and lanes using a vision system on the vehicle. Detecting all kinds of obstacles on the road, mainly include the IPM (Inverse Perspective Mapping) method. The system acquires the front view using a camera mounted on the vehicle then applying a few processes in order to detect the lanes and objects. A versatile methodology is used in order to detect the lanes and objects. In our research we have developed a simple heuristic method which is more robust in both lane detection object detection and tracking in video. In this method we use clustering methodology to group the detected points in case of lane detection. Heuristic gives effective results in detection and tracking of multiple vehicles at a time irrespective to the distance.

Distributed event-triggered adaptive output-feedback formation tracking of uncertain underactuated underwater vehicles in three-dimensional space

We present an adaptive-observer-based event-triggered design for distributed three-dimensional output-feedback formation tracking of multiple uncertain underactuated underwater vehicles (UUVs) under a directed network. It is assumed that all nonlinear functions of the vehicle dynamics are unknown and the velocity information of vehicles is immeasurable for feedback. Compared with existing distributed tracking results for UUVs, the primary contribution of this study is the development of a state-transformation-based adaptive observer using a time-varying scaling signal to achieve the output-feedback three-dimensional formation tracking in the presence of unknown nonlinearities and the underactuation constraint. Based on the state transformation, local adaptive observers using neural network approximators and their adaptive laws are designed to estimate local unmeasured velocities of UUV followers. By designing distributed nonlinear error functions and auxiliary stabilizing signals, local output-feedback tracking laws with triggering conditions are recursively constructed to ensure preselected formation performance while solving the underactuation problem of the UUV dynamics. It is shown that the formation tracking errors remain within pre-designable time-varying bounds and finally converge to a neighborhood of the origin. Simulation results are presented to show the effectiveness of the suggested output-feedback formation tracking strategy.

Multiple Pedestrians and Vehicles Tracking in Aerial Imagery Using a Convolutional Neural Network

In this paper, we address various challenges in multi-pedestrian and vehicle tracking in high-resolution aerial imagery by intensive evaluation of a number of traditional and Deep Learning based Single- and Multi-Object Tracking methods. We also describe our proposed Deep Learning based Multi-Object Tracking method AerialMPTNet that fuses appearance, temporal, and graphical information using a Siamese Neural Network, a Long Short-Term Memory, and a Graph Convolutional Neural Network module for more accurate and stable tracking. Moreover, we investigate the influence of the Squeeze-and-Excitation layers and Online Hard Example Mining on the performance of AerialMPTNet. To the best of our knowledge, we are the first to use these two for regression-based Multi-Object Tracking. Additionally, we studied and compared the L1 and Huber loss functions. In our experiments, we extensively evaluate AerialMPTNet on three aerial Multi-Object Tracking datasets, namely AerialMPT and KIT AIS pedestrian and vehicle datasets. Qualitative and quantitative results show that AerialMPTNet outperforms all previous methods for the pedestrian datasets and achieves competitive results for the vehicle dataset. In addition, Long Short-Term Memory and Graph Convolutional Neural Network modules enhance the tracking performance. Moreover, using Squeeze-and-Excitation and Online Hard Example Mining significantly helps for some cases while degrades the results for other cases. In addition, according to the results, L1 yields better results with respect to Huber loss for most of the scenarios. The presented results provide a deep insight into challenges and opportunities of the aerial Multi-Object Tracking domain, paving the way for future research.

Multi‐camera multiple vehicle tracking in urban intersections based on multilayer graphs

Vehicle visual tracking is a challenging issue in intelligent transportation systems. The tracking gets more challenging when vehicles change direction at intersections. Undetermined motion flows, occlusion, and congestion are the potential issues of vehicle tracking at intersections. In this study, a new method for tracking multiple vehicles from a multi-view is proposed to overcome occlusion caused at the intersections with undetermined motion flows. In the authors’ method, a multilayer graph is presented that assigns motion flows to distinct layers with different neighbourhoods for each layer represented by the graph's edges. Hence, the vehicle trajectories are distributed among layers such that vehicles entering from the same side with similar motion flows are assigned to the same layer. All multilayer graphs of different views are mapped to the graph of the selected view. Then, tracking is performed on the distinct layers of the mapped multilayer graph by computing min-cost flows. In cases such as vehicle crossing, misdetection, or occlusion, the method can predict the vehicle's tracks by using history, layer neighbourhoods, and other views’ information. Experimental results show a consistency of the ground truth and the analysis obtained using the proposed method in tracking vehicles in the inner part of the intersection.

Multi-UAV cooperative target tracking with bounded noise for connectivity preservation

We investigate cooperative target tracking of multiple unmanned aerial vehicles (UAVs) with a limited communication range. This is an integration of UAV motion control, target state estimation, and network topology control. We first present the communication topology and basic notations for network connectivity, and introduce the distributed Kalman consensus filter. Then, convergence and boundedness of the estimation errors using the filter are analyzed, and potential functions are proposed for communication link maintenance and collision avoidance. By taking stable target tracking into account, a distributed potential function based UAV motion controller is discussed. Since only the estimation of the target state rather than the state itself is available for UAV motion control and UAV motion can also affect the accuracy of state estimation, it is clear that the UAV motion control and target state estimation are coupled. Finally, the stability and convergence properties of the coupled system under bounded noise are analyzed in detail and demonstrated by simulations.

An intelligent multiple vehicle detection and tracking using modified vibe algorithm and deep learning algorithm

Multiple vehicle detection is a promising and challenging role in intelligent transportation systems and computer vision applications. Most existing methods detect vehicles with bounding box representation and fail to offer the location of vehicles. However, the location information is vigorous for several real-time applications such as the motion estimation and trajectory of vehicles moving on the road. In this paper, we propose an advanced deep learning method called enhanced you only look once v3 and improved visual background extractor algorithms are used to detect the multi-type and multiple vehicles in an input video. More precisely, tracking is to find the trace of the upcoming vehicles using a combined Kalman filtering algorithm and particle filter techniques. To improve the tracking results, further, we propose the technique, namely multiple vehicle tracking algorithms, and tested with different weather conditions such as sunny, rainy, night and fog in input videos of 30 frames per second. The major research issues were found in the recent kinds of literature in ITS sector which is closely related to the real-time traffic environmental problems such as occlusions, camera oscillations, background changes, sensors, cluttering, camouflage, varying illumination changes in a day- and sunny and at nighttime vision. The experimental results are tested with the ten different input videos and two benchmark datasets KITTI and DETRAC. The most eight high- level features have been considered for automatic feature extraction and annotation. The attributes are length, width, height, number of mirrors and wheels and windscreen shielding glass to detect the target region of interest (vehicles) on road. In addition, further experiments are carried out in multiple-input videos of high definition quality using a monocular camera, and the average accuracy is 98.6%, and the time complexity of the algorithm is O(n) and also tracking results attained 96.6%. The dataset and input videos are discussed in comparative results with the F-test measure done for multiple vehicles.