The application scenarios of automated robots are undergoing a paradigm shift from structured environments to unstructured, complex settings. In highly constrained settings like factory inspections or disaster rescue, conventional steering systems show clear drawbacks. While the four-wheel independent drive and independent steering (4WS) robot provides a variety of steering modes, which can effectively meet the needs of complex environments. However, how a 4WS robot autonomously selects different steering modes based on trajectory point information during trajectory tracking remains a challenging problem. This paper proposes a multi-modal trajectory tracking method considering the switch of steering modes, which decomposes the trajectory tracking task into two parts: mode decision-making and tracking control. The corresponding method is designed based on deep reinforcement learning. Additionally, a target trajectory random generator and corresponding training interaction environment are designed to train the model in a data-driven manner. In the designed scenario, our tracker achieve more than a 30% improvement in average tracking error across all motion modes compared with model predictive control, and the decider’s average decision position error is less than 2 cm. Extensive experiments demonstrate that our method achieves superior tracking performance and real-time capabilities compared to current methods.

Regular temperature measurement of critical parts of an annealing furnace has always been a difficult task. Due to the harsh environment of high temperature, high noise, and darkness in the annealing furnace operation area, unmanned vehicles equipped with the RGB-T semantic segmentation model are usually adopted in most factories for inspection. However, existing RGB-T semantic segmentation models usually rely on good lighting or thermal conditions, which are generally difficult to fulfill in annealing furnace operation areas. In this paper, we propose a new hierarchical fusion-based semantic enhancement network, HFSENet. We first adopt the two-stream structure and the siamese structure to extract the low-level and high-level features of unimodal modalities, respectively. Then, considering the differences between the features in different hierarchical levels, we introduce a novel low-level feature spatial fusion module and a high-level feature channel fusion module to perform the multi-modal feature hierarchical fusion. On this basis, we also propose the semantic feature complementary enhancement module, which utilizes the appearance information set and object information set extracted from RGB and thermal infrared (TIR) branches to enhance the fused features and give them more semantic information. Finally, segmentation results with refined edges are obtained by an edge refinement decoder that includes a local search extraction module. The unmanned inspection vehicle we built with the proposed HFSENet has successfully passed the test, and the recognition performance of the four targets exceeds the current state-of-the-art (SOTA) method on our homemade annealing furnace operation area dataset.

With the rapid advancement of intelligent manufacturing and the rise of emerging markets, global auto-mobile exports have surged, placing unprecedented demands on logistics infrastructure. Efficient coordination of multiple robots for vehicle autonomous transfer is essential in high-density storage environments. However, conventional navigation mode, where autonomous robots navigate the entire space, often leads to inefficiencies, congestion, and increased safety risks. To address these challenges, this paper proposes a dynamic network topology framework to optimize large-scale vehicle transfers in high-density environments. The approach models free space as a network graph with directional, weighted movement costs. Leveraging yard operational characteristics, real-time transfer conditions, and robot specific capabilities, we introduce an event-triggered mechanism to update the network topology dynamically. This method continuously refines drivable space, effectively integrating yard areas with roadways to enhance routing flexibility in robot scheduling. Scenario-Based evaluations demonstrate that the proposed approach reduces traveled distance by up to 12.3% and task completion time by 19.3% compared to traditional operational networks, leading to lower operational costs and improved task efficiency. Notably, these benefits become more pronounced as the number of robots increases and the operational environment grows more complex.

Modern autonomous systems face significant challenges in coordinating multiple agents within complex environments, especially under non-holonomic constraints and kinodynamic motion planning demands. In particular, four-wheel independent drive/steering (4WIDS) robots, with their multimodal locomotion and non-holonomic kinematics, introduce new complexity to multi-agent pathfinding problems in continuous space. This study addresses the critical challenge of generating collision-free, dynamically feasible paths for multi-4WIDS robot systems. We propose a hierarchical MAPF framework that explicitly tackles two core limitations in existing methods: (1) a lack of physically realistic inter-agent conflict modeling, and (2) the inability to account for motion mode switching costs in kinodynamic path planning. At the upper level, we introduce a continuous-space binary conflict tree for resolving space-time collisions using physical dimensions and motion feasibility. At the lower level, we implement a hybrid A* search over a motion primitive database that considers mode-switch penalties and kinematic constraints. To improve search performance, we introduce two novel mechanisms: a focused search factor for directional exploration and an adaptive heuristic weighting factor to balance optimality and computational speed. Extensive experiments, including benchmark comparisons, ablation studies, and sensitivity analysis, validate the proposed method’s superior performance in inter-agent conflict resolution, real-time feasibility, and path quality.

In dynamic and unstructured environments, mobile robots need to generate safe and efficient trajectories in real time, which poses significant challenges due to the uncertainty of surrounding obstacles. To address this, this article presents a real-time obstacle avoidance trajectory planning method, built upon a spatio-temporal graph network that integrates temporal modeling with graph attention mechanisms. The proposed network captures both temporal dynamics and spatial structural dependencies in dynamic environments by integrating a temporal information module based on long short-term memory (LSTM) and a spatial module based on relational graph attention networks (RGAT). On the whole, the approach follows a two-phase pipeline. In the offline phase, a high-quality trajectory dataset is constructed to represent the heterogeneous state graph of the robot and surrounding obstacles. Then the dataset is used to train the spatio-temporal network, which learns to map environment-state graphs to optimal control commands. In the online phase, the trained network is deployed on the robot to perform real-time perception, decision-making, and control, forming a closed-loop trajectory optimization process. Extensive experiments in both simulated and real-world scenarios demonstrate that the proposed method achieves high-quality trajectory planning, robust obstacle avoidance, and fast generalization under multi-obstacle and sudden disturbance conditions, while maintaining low computational overhead.

Hierarchical torque vector control has proven effective in enhancing stability and handling in vehicles; however, its application in mobile robots faces challenges due to dynamic model variations and hardware differences. In response, this paper introduces a novel framework that leverages rich sensor information to optimize torque vector control in mobile robots. In this paper, we establish a comprehensive total torque parameter-varying error (TTPE) model at the upper level, effectively decoupling the yaw moment generated by the steering angle from the additional yaw moment. To achieve accurate and stable control, we employ a TTPE model predictive controller (TTPE-MPC) that generates precise desired total torque. Furthermore, we introduce a slip loss objective function based on kinematics in the middle layer, in addition to the conventional torque error and tire load objective functions. We enhance the optimization of tire load by incorporating information on the six-dimensional leg force, which significantly improves maneuverability and tracking precision during the torque distribution stage. To further enhance precision and real-time performance of torque distribution, a disturbance observer is designed for each individual wheel to estimate the tire’s longitudinal force at the lower layer. The control law of the motor input current is then designed based on this estimation, resulting in improved control accuracy. The proposed framework is validated through experiments conducted on a six-wheeled robot. The results demonstrate that the stability of the expected value of the total moment improves, with tracking errors of velocity and angular velocity reduced by approximately 43%. Moreover, the slip loss between wheels is reduced by approximately 46% compared to conventional methods.Note to Practitioners—This article aims to improve robotic steering maneuverability and provide safer, stable solutions for autonomous vehicles with independent propulsion. Current methods for robot operation often lack optimal torque management, primarily focusing on direct speed control while neglecting valuable sensor data. We propose a model-based, hierarchical control strategy to efficiently allocate torque to the robot’s propulsion motors. Our approach establishes a model that distinguishes between angular difference and motor torque, employing a Total Torque Parameter-Varying Error Model Predictive Controller (TTPE-MPC) for cumulative torque calculation. Optimization metrics, with a focus on sensor data, guide torque allocation, taking anti-slip considerations into account. Furthermore, a speed-based control observer ensures fair torque distribution, theoretically converging expected, estimated, and true values. Initial robot experiments validate reduced speed and angular velocity errors, supporting the effectiveness of our approach. Future work will explore incorporating suspension system aerodynamics into torque distribution, further enhancing stability.

In the context of unstructured and unknown environment, the autonomous navigation still faces many challenges, such as assessing rough terrain and deciding how to safely navigate complex terrain. In this work, we propose a robust and practical off-road navigation framework that has been successfully deployed on a vibroseis truck for land exploration. First, in degraded wild scenes, a tightly coupled lidar-GNSS-inertial fusion odometry and mapping framework is adopted to construct a local point cloud map around the vehicle in real-time and provide precise localization. Then, based on amplitude-frequency characteristic analysis and point cloud PCA, a multi-layer terrain assessment map containing terrain roughness, obstacles and slope information is obtained. Finally, combining Gaussian distribution based adaptive sampler and Bayesian sequentially updated proposal distribution, a local graph is efficiently built to obtain multiple path solutions under constrained conditions. Both simulations and field experiments show that the proposed navigation framework can decide how to travel on a flat road even in harsh terrain conditions, naturally suppressing frequent attitude angle changes and preventing vehicle accidents.

In order to guarantee autonomous vehicles' autonomy, mobility, and ride quality in rugged environments, a real-time speed planning method based on the time-frequency transformation of terrain characteristics is designed to achieve adaptive speed planning of autonomous vehicles in rough ground. On the one hand, the vertical profile of the lidar's point cloud data is converted from the time domain to the frequency domain in real time, and the integrated area of the sub-frequency range in the frequency domain is chosen as the relative roughness quantification value to realize the roughness quantification under various terrains. On the other hand, to model the relationship between vehicle speed and relative roughness, iterative search is utilized to create a speed and roughness model, and sliding windows are employed to update the roughness to achieve continuous mapping between speed and roughness. Ultimately, a number of tests were conducted on various rough roads using the oil exploration vehicle EV-56 as the study object. The experimental results show that the proposed method can identify the terrain roughness changes under complex terrain and change their speed within 0.2 m accuracy.