Security robots often operate in environments characterized by low light and smoke, where millimeter-wave radar proves effective. However, the millimeter-wave radar’s point cloud is often sparse and noisy, potentially leading to positioning failure when employing point cloud matching. In this paper, we propose a localization strategy for security robots based on millimeter wave radar. The position of the security robot is deduced by clustering and tracking the sparse point cloud. In addition, radial velocity was used to design an adaptive tracking threshold, so that the targets in the two frames of data are within the tracking threshold regardless of the fast or slow motion speed of the security robot. Experimental results indicate that this method circumvents positioning failures associated with point cloud matching. In comparison to the radial velocity method, this approach enhances positioning accuracy by approximately 33.9%. Additionally, compared to the fixed tracking association threshold, this method exhibits a higher success rate in target tracking, leading to a 25.9% improvement in security robot positioning accuracy.

In smoggy and dusty environments, vision- and laser-based localization methods are not able to be used effectively for controlling the movement of a robot. Autonomous operation of a security robot can be achieved in such environments by using millimeter wave (MMW) radar for the localization system. In this study, an approximate center method under a sparse point cloud is proposed, and a security robot localization system based on millimeter wave radar is constructed. To improve the localization accuracy of the robot, inertial localization of the robot is integrated with MMW radar. Based on the concept of inertial localization, the state equation for the motion principle of the robot is deduced. According to principle of MMW localization, the measurement equation is derived, and a kinematics model of the robot is constructed. Further, by applying the Kalman filtering algorithm, a fusion localization system of the robot based on MMWs and inertial localization is proposed. The experimental results show that with iterations of the filtering algorithm, the gain matrix converges gradually, and the error of the fusion localization system decreases, leading to the stable operation of the robot. Compared to the localization system with only MMW radar, the average localization error is approximately reduced from 11 cm to 8 cm, indicating that the fusion localization system has better localization accuracy.

In modern urban logistics and schedule systems, road congestion stands out as a primary contributor to heightened energy consumption in new energy logistics vehicles. Addressing this issue, this study establishes a scheduling method for new energy logistics vehicles comprising several key components: Using the vehicle–road–cloud collaborative technology, the number of vehicles on the road is obtained, and the road congestion coefficient is calculated by combining the speed‐flow model, and then the nonlinear energy consumption model for new energy logistics vehicles is studied. Additionally, a VRC‐GVRP model is developed considering multiple constraints with the aim of minimizing total energy consumption. To solve this model, an initial solution is constructed using an energy‐saving algorithm, while exploring a Cauchy variational strategy and a parallel local search to propose an improved adaptive large neighborhood search (ALNS) algorithm. An illustrative analysis is conducted within an industrial park, based on the real‐time traffic information aggregated to the cloud control platform, and the scheduling problem of new energy logistics vehicles is solved. The experimental results indicate that the enhanced ALNS algorithm exhibits rapid convergence and yields high‐quality solutions. Compared to the situation without vehicle–road–cloud collaboration technology, despite the increase in the total distance traveled by new energy logistics vehicles, the proposed method effectively reduces total drive time and total energy consumption. As the congestion factor increases, the percentage of reduction in total time and total energy consumption becomes higher and higher, indicating that this method is of great significance for improving the work efficiency of new energy logistics vehicles and achieving energy conservation and emission reduction.

It is difficult for a robot to capture object which is thrown out in the environment with natural wind. Equation of state and measurement of the object in the environment with natural wind is studied. Using the measurement equation and the state equation when there is no natural wind, the average acceleration of the object caused by natural wind can be estimated. Moreover, the variance of the object acceleration caused by natural wind can be estimated using the statistical characteristics and probability distribution of natural wind. Then a method for robot capturing the object based on Kalman filter is proposed. Simulation results show that the trajectory of the object changes obviously when two groups of typical natural wind are used. The object trajectory can be tracked by the robot accurately after it deviates from the object trajectory for about eight sampling periods.

Dynamic traffic flows at unsignalized intersections generate complex interaction cascades among different traffic participants, where indirect behavioral influences propa gate through inter-agent networks, thereby compromising autonomous vehicle safety and operational efficiency. To address the limitations of multi-level interaction perception caused by interactive propagation between traffic participants at unsignalized intersections, this paper proposes a graph attention network integrated deep reinforcement learning (GAT-DRL) framework for autonomous driving decision-making. The model constructs an interaction topology network and dynamic traffic graph using spatiotemporal states of ego/surrounding vehicles and road structural features. A spatial perception module leveraging multi head graph attention network is designed to capture global interaction features through integrated risk-aware strategies. The global risk weight matrix, produced by the graph attention network, is tensor-concatenated with vehicle kinematic states and serves as input to the Soft Actor-Critic (SAC) module for policy optimization. Extensive experiments demonstrate that the proposed model achieves superior performance across multiple evaluation metrics, including cumulative reward, success rate, and average completion time, outperforming baseline reinforce ment learning methods.

In this paper, a novel target tracking and filtering method is proposed to address the challenges of millimeter-wave radar perception in industrial environments, including multipath interference, high noise levels, and stringent real-time requirements. A velocity-constrained multidimensional Density-Based Spatial Clustering of Applications with Noise (DBSCAN) clustering algorithm is developed to improve clustering accuracy and stability by exploiting the triangle inequality to accelerate processing and reduce false alarms. For target tracking, a constant-velocity (CV) motion model is adopted, and radial velocity information is incorporated to derive a nonlinear measurement equation for constructing an extended Kalman filter (EKF) framework. Experimental results demonstrate that the proposed method improves clustering accuracy by 19%, achieves 94% clustering stability, and increases clustering speed by approximately eight times. Additionally, compared with the DBSCAN+KF-based tracking method and the GTRACK method, the proposed method achieves a higher tracking success rate and more accurate robot tracking.