Journal of Field Robotics最新文献

This paper presents the development of a Physics-realistic and Photo-realistic humanoid robot testbed, PR2, to facilitate collaborative research between Embodied Artificial Intelligence (Embodied AI) and robotics. PR2 offers high-quality scene rendering and robot dynamic simulation, enabling (i) the creation of diverse scenes using various digital assets, (ii) the integration of advanced perception or foundation models, and (iii) the implementation of planning and control algorithms for dynamic humanoid robot behaviors based on environmental feedback. The beta version of PR2 has been deployed for the simulation track of a nationwide full-size humanoid robot competition for college students, attracting 137 teams and over 400 participants within 4 months. This competition covered traditional tasks in bipedal walking, as well as novel challenges in loco-manipulation and language-instruction-based object search, marking a first for public college robotics competitions. A retrospective analysis of the competition suggests that future events should emphasize the integration of locomotion with manipulation and perception. By making the PR2 testbed publicly available at https://github.com/pr2-humanoid/PR2-Platform, we aim to further advance education and training in humanoid robotics. Video demonstration: https://pr2-humanoid.github.io/.

The articulated quad-track vehicles exhibit excellent mobility and obstacle-crossing capabilities in outdoor environments, making them widely applicable in agriculture and military fields. Their steering control is a complex issue influenced by numerous factors. To reduce the computational complexity of the controller, achieve rapid system response, and simultaneously improve the stability and precision of the articulated quad-track vehicles during the steering control process, an optimal matching analysis is performed between the inner and outer track speed ratios and the deflection angles at the front and rear articulation points of the vehicle. By utilizing fuzzy proportional–integral–derivative control and visual navigation, a path-tracking control experimental platform for the articulated quad-track vehicle is designed. Through a combination of virtual prototype simulations and physical experiments, the distance deviation and heading angle deviation between the actual driving path of the virtual and experimental prototypes and the preset path are analyzed. The designed path–tracking control system can adjust the driving speeds of the left and right tracks and the articulation point deflection angle based on the preset driving path, enabling the vehicle to track the path. Under stable steering conditions, the distance deviation is within 0.1 m, and the heading angle deviation is within 6°, demonstrating excellent control performance.

In the dynamic field of agricultural technology, the development of coconut tree climbers exemplifies significant progress in addressing the challenges of efficient and safe coconut harvesting. Designing an unmanned coconut tree climber robot is complex due to the unpredictable structures of the coconut tree trunk and crown. Key challenges include developing a climbing mechanism, ensuring smooth ascents and descents, managing payload stability, and designing an effective harvester for coconut bunches, all of which impact the robot's overall performance. This paper introduces a novel design featuring a double-ring structure for the climber robot, aimed at enhancing its performance. The study includes a comprehensive static analysis to determine the average range of torque values for the actuators. Dynamic and kinematic analyses are conducted to establish essential relationships that predict the robot's characteristics before testing. A four-degree-of-freedom manipulator is used as the harvester. The proposed methodology was tested on a coconut tree trunk in a lab setup and field conditions across 10 different coconut trees. Real-time data collected during these tests were validated against predictions made through simulations before experimentation. The analyses, including theoretical analysis, simulation outcomes, and experimental test setups, conclusively demonstrate that the proposed structure maintains consistent stability throughout the climbing process, even on trees with varying inclinations and trunk radii relative to height. The success rates of the double-ring setup consistently surpass those of the single-ring configuration, with success rates ranging from 80% to 100% for the single ring and 100% for the double-ring setup.

Due to the growing complexity of diverse maritime tasks, underactuated unmanned surface vehicle (USV) has become a research hotspot. The rapid development of deep reinforcement learning (DRL) technology has brought forth a novel approach for the USV autonomous control, rendering unnecessary the dynamical modeling of the target USV. To further improve the USV collision avoidance performance against maritime disturbances, this paper presents a predictive reinforcement learning method for USV obstacle avoidance control. A prediction module is designed to generate latent features that depict environmental states. After that, the prediction feature is provided for a DRL-based policy module to produce an action distribution for the underactuated unmanned surface vehicle. The proposed method in this paper can enable the USV avoid obstacle and reach the destination solely based on its local observational information, without relying on prior global information. Simulation and physical experiments have demonstrated that, compared to general DRL methods, the proposed method exhibits stronger robustness to environmental disturbances, enabling the USV to reach the destination while avoid the obstacle.

Mobile robot route planning is the process of calculating a mobile robot's collision-free path from a starting place to a goal point surrounded by its environment. It is a critical component of mobile robotics since it enables robots to move around and perform tasks independently in a variety of conditions. The Global Positioning System (GPS), the Adaptive Neuro-Fuzzy Inference System (ANFIS), and deep reinforcement learning (DRL) are commonly used tools for tracking as well as control. This paper proposes a GPS-based DRL-ANFIS navigation method for mobile robots that avoid collisions. The GPS-based controller keeps the robot on track to achieve its global and flexible objective. Next, a fuzzy inference system (FIS) is employed to simulate obstacle avoidance using fuzzy linguistics on distance sensor data. In addition, a mobile robot path planning technique based on enhanced DRL is proposed to address the issues of limited exploration capability and sparse reward of environmental state space in mobile robot route planning in unfamiliar environments. Finally, the proposed ANFIS parameters are fine-tuned using a tent-based artificial hummingbird algorithm (AHA) to attain the desired location. The proposed approach evaluates the results using MATLAB. The simulation study is designed to assess the proposed strategy's effectiveness in navigating a mobile robot across a complex environment, as well as its performance in comparison to existing collision-free navigation systems. As a result, the proposed approach takes a shorter path and avoids barriers to get the robot closer to its destination. The proposed approach has a computation time of 22 s and a path planning efficiency of 96.56%, which is 5.56% higher than the traditional DRL model.

Legged robots provide an efficient alternative for navigation in complex terrains. However, few studies have explored dynamic locomotion for hexapod robots navigating unknown, rugged terrains. In this paper, the design, control, and implementation of a hexapod robot, RENS H3, inspired by the lateral movement of crabs, are presented with a focus on its adaptability in unknown, uneven terrains. The robot's structural design is introduced, and a hardware control framework for hexapod robots is developed. Additionally, a hierarchical control framework based on model predictive control is proposed, integrating terrain-adaptive control and foot-end Cartesian space force compensation based on posture adjustment into the control architecture to enhance the robot's robustness and terrain adaptability on slopes and unstructured terrains. The proposed method's robustness, adaptability, and energy efficiency were demonstrated through a series of experiments conducted on various outdoor slope terrains, unstructured terrains, and the multi-terrain testbed. Comparative experimental tests further validated the advantages of the approach in unknown rugged terrains.

Trajectory tracking control is a fundamental problem in the control of unmanned systems. In practical systems, actuators often have input quantization and saturation constraints, and failing to account for these constraints can affect control convergence time, precision, and even lead to system instability. Therefore, designing a practical predefined time controller specifically for unmanned systems with input quantization and saturation is particularly important. In this study, a novel practical predefined-time control criterion and predefined-time control criterion are proposed. A novel observer with predefined-time convergence is designed which can deal with both input quantization and input saturation. It is efficient without high computational cost, local optima, or complex parameter tuning. It can deal with ship systems with 6 degrees of freedom exposed to external disturbance. The 6 degrees of freedom model provides a more comprehensive representation of the dynamic characteristics. An autonomous vehicle model is used for testing, and the effectiveness of the proposed algorithm has been demonstrated.