Journal of Field Robotics最新文献

This study presents a novel method for analyzing the motion path of a differential drive mobile robot using a single video camera and perspective transformation. The robot utilized is the Zoom:Bit by Cytron Technologies, with motion planning based on clothoids, specifically the C¹-Continuous Double Clothoid Segments introduced by Idris et al. The robot's movement is generated by calculating wheel speeds along the clothoid path, with the technical configuration also discussed in detail. In contrast to conventional methods that depend on automatic tracking systems such as GPS and sensor fusion, this approach adopts a cost-effective setup using a single camera mounted on a tripod to capture the robot's motion. The recorded video is processed using Mathematica, where keyframes are extracted, and the robot's coordinates are automatically identified at specific time intervals. Perspective transformation is then employed to convert the recorded 3D motion into a 2D plane, enabling a detailed comparison between the robot's actual trajectories and simulated results. The findings underscore the feasibility and accuracy of using a single-camera system for 3D motion capture and validate the efficiency of C¹-Continuous Double Clothoid Segments for motion planning, offering a practical and cost-efficient solution for motion path analysis in mobile robotics.

A cooperative formation control strategy for multiple autonomous underwater vehicles (AUVs) is proposed for underwater topography detection and validated through field trials with the Wukong 1000 AUV. The strategy incorporates a leader-follower communication structure, which defines both communication topology and timing. To enhance path-following accuracy, an improved guidance law is developed for the leader AUV. Additionally, a state estimation method is introduced to mitigate the effects of communication delays and intervals, enabling follower AUVs to accurately predict the leader's real-time state and ensuring formation integrity. A motion controller for underactuated systems is integrated with the guidance law to support precise trajectory tracking. To maintain range consistency and ensure safe, stable maneuvering, a cooperative steering protocol based on equivalent travel distance is also proposed. The method's effectiveness and robustness are validated through extensive field experiments using a tri-AUV system. These experiments demonstrate the strategy's ability to conduct large-scale, prolonged underwater topography detection operations with high operational efficiency and safety. Notably, even under challenging conditions—such as a communication interval exceeding 14 s, communication delays greater than 2 s, and a communication success rate of nearly 90%—the formation tracking error remains under 2 m during cooperative seabed exploration.

A single-beacon, ranging-based acoustic guidance method is proposed to ensure reliable, rapid, and omnidirectional homing guidance for the recovery of autonomous underwater vehicle (AUV). To address the navigation system deviations caused by prolonged underwater operations of both the AUV and the mothership, the method first unifies their navigation datums. It then uses limited moving target information and ranging data to estimate the position of the moving target, enabling effective AUV recovery and homing guidance. This method is theoretically applicable across a wide range of distances—from tens of meters to several kilometers—and is well-suited for autonomous homing. It was adopted for its ability to overcome the key limitations of traditional ultrashort baseline systems, which are constrained by non-omnidirectional localization and susceptibility to positional errors resulting from relative positioning between the AUV and the mothership. Results from trials at the Danjiangkou Reservoir demonstrate, for the first time, that an AUV can be accurately guided to the vicinity of the mothership using measurement data from a single acoustic communication device. To evaluate the adaptability of the proposed acoustic guidance algorithm under large initial position errors, 14 field experiments were conducted with the mothership operating under both dynamic positioning (DP) and moving conditions. The results showed that homing deviation was consistently maintained within 8 m, with best-case values of 2.6 m (DP) and 3.5 m (moving), and corresponding average deviations of 5.8 and 6.5 m. These findings confirm the robustness and effectiveness of the proposed method under challenging initial deviation scenarios.

Object detection under low-light conditions is a critical perceptual capability for autonomous robotic systems performing real-world tasks such as autonomous driving, nighttime surveillance, drone inspection, and underwater exploration. However, image degradation in such environments severely disrupts visual feature extraction, leading to frequent false positives and missed detections. To address this challenge, we propose a novel Multi-scale Adaptive Network (MANet) for robust object detection in low-light scenarios. MANet comprises two main components: a cascaded feature extraction network built upon our proposed multi-scale feature extractor, and an adaptive fusion network that integrates our adaptive feature extractor and a fast normalized fusion module. Additionally, we introduce a joint loss function to further improve classification performance in complex lighting conditions. Experimental results show that MANet achieves an mAp@0.5 of 0.718 and an mAp@0.5:0.95 of 0.451 on the ExDark data set, and also delivers competitive performance on DARKFACE, DUO, and TrashCan. In addition, MANet demonstrates strong cross-scene generalization under real-world low-light conditions. These results validate the effectiveness of MANet in reducing false positives and missed detections, enhancing detection robustness, and laying a solid foundation for robotic perception and decision-making in real-world outdoor low-light environments.

We present a comprehensive evaluation of a point-cloud-based navigation stack, MUONS, for autonomous off-road navigation. Performance is characterized by analyzing the results of 30,000 planning and navigation trials in simulation and validated through field testing. Our simulation campaign considers three kinematically challenging terrain maps and twenty combinations of seven path-planning parameters. In simulation, our MUONS-equipped AGV achieved a 0.98 success rate and experienced no failures in the field. By statistical and correlation analysis, we determined that the Bi-RRT expansion radius used in the initial planning stages is most correlated with performance in terms of planning time and traversed path length. Finally, we observed that the proportional variation due to changes in the tuning parameters is remarkably well correlated to performance in field testing. This finding supports the use of Monte-Carlo simulation campaigns for performance assessment and parameter tuning.

In robot perception, convolutional networks are representative of hierarchical neural networks and have achieved outstanding results in fields of text and image processing. The key to extracting local features and integrating them into global features is the presence of shared regions between windows, which is a significant factor in the success of convolutional neural networks. Inspired by this success, this paper investigates existing methods on hierarchical point cloud network methods for grouping point clouds and proposes a solution involving shared anchor points to enable subsets to share points. This effectively enhances the feature extraction capability of existing hierarchical point cloud networks. Building upon this, we have designed a slice-based anchor point transformer called SAPT, which exhibits strong compatibility with downstream tasks. Experimental results demonstrate that, with assistance of shared anchor points, extracted features are not only suitable for classification and segmentation tasks but also perform exceptionally well in object detection tasks, without the need for designing specific network enhancements for local feature extraction. Empirical evidence shows that SAPT excels in both classification and segmentation as well as object detection tasks. On KITTI (LiDAR only), with a threshold of 0.7 IoU and 40 recall points, the mAP value reached a maximum of 79.85.

The ocean's vast resources drive exploration interest, yet its complex and dynamic nature demands advanced underwater robotic systems. This paper presents an autonomous docking system utilizing a bionic vector-propelled robotic fish. The robotic fish, driven by a wire-actuated vector-propulsion tail, achieves omnidirectional movement. Inspired by biological swarm coordination, a bidirectional information exchange protocol is developed for the docking process. To enhance autonomy, we design a hierarchical closed-loop control system integrating a central pattern generator (CPG), visual recognition, and serial controllers. This enables the robotic fish to complete release, search, task execution, and docking. Experimental validation demonstrates the system's effectiveness in achieving stable docking and coordinated motion. The results confirm the feasibility and reliability of the proposed methodology, offering a novel solution for underwater docking in dynamic environments.

With the development of soft crawling robots, pneumatic soft actuators (PSAs) with complicated nonplanar structures are increasingly designed due to the capabilities of achieving compound robot movements. However, the current deformation accuracy of complex PSAs is insufficient to satisfy application requirements. Specifically, the dynamic properties of complex PSAs are still ambiguous and PSA deformation control methods remain lacking, which are caused by inherent characteristics, such as strong nonlinearity and hysteresis. To this end, based on self-fabricated spatial deformation pneumatic leg actuators (SDPLAs), an equivalent dynamic model derived from Euler–Lagrange equations and a modified hysteresis model are established to form the new complete model of SDPLA systems, which accurately describes the bending deformation and hysteresis of SDPLAs at the same time. Furthermore, a novel model-based adaptive fuzzy tracking controller is designed for SDPLAs, which addresses model uncertainties and unknown external torque, and achieves the accurate bending of each SDPLA part. Subsequently, the closed-loop stability is rigorously proven by the Lyapunov theory. Finally, a series of experiments validates the effectiveness of the established dynamic and hysteresis models, the tracking control performance on time-varying trajectories with different amplitudes, frequencies, and shapes, and the control robustness against disturbances.