Journal of Field Robotics最新文献

LiDAR-based 3D place recognition is an essential component of simultaneous localization and mapping systems in multi-scene robotic applications. However, extracting discriminative and generalizable global descriptors of point clouds is still an open issue due to the insufficient use of the information contained in the LiDAR scans in existing approaches. In this paper, we propose a novel spatial-temporal point cloud encoding network for multiple scenes, dubbed STM-Net, to fully fuse the multi-view spatial information and temporal information of LiDAR point clouds. Specifically, we first develop a spatial feature encoding module consisting of the single-view transformer and multi-view transformer. The module learns the correlation both within a single view and between two views by utilizing the multi-layer range images generated by spherical projection and multi-layer bird's eye view images generated by top-down projection. Then in the temporal feature encoding module, we exploit the temporal transformer to mine the temporal information in the sequential point clouds, and a NetVLAD layer is applied to aggregate features and generate sub-descriptors. Furthermore, we use a GeM pooling layer to fuse more information along the time dimension for the final global descriptors. Extensive experiments conducted on unmanned ground/surface vehicles with different LiDAR configurations indicate that our method (1) achieves superior place recognition performance than state-of-the-art algorithms, (2) generalizes well to diverse sceneries, (3) is robust to viewpoint changes, (4) can operate in real-time, demonstrating the effectiveness and satisfactory capability of the proposed approach and highlighting its promising applications in multi-scene place recognition tasks.

Soft robots face significant challenges in proprioceptive sensing and precise control due to their highly deformable and compliant nature. This paper addresses these challenges by developing a conductive hydrogel sensor and integrating it into a soft robot for bending angle measurement and motion control. A quantitative mapping between the hydrogel resistance and the robot's bending gesture is formulated. Furthermore, a nonlinear differentiator is proposed to estimate the angular velocity for closed-loop control, eliminating the reliance on conventional sensors. Meanwhile, a controller is designed to track both structural and nonstructural trajectories. The proposed approach integrates advanced soft sensing materials and intelligent control algorithms, significantly improving the proprioception and motion accuracy of soft robots. This work bridges the gap between novel material design and practical control applications, opening up new possibilities for soft robots to perform delicate tasks in various fields. The experimental results demonstrate the effectiveness of the proposed sensing and control approach in achieving precise and robust motion control of the soft robot.

Over the last few decades, the agricultural industry has made significant advances in autonomous systems, such as wheeled robots, with the primary objective of improving efficiency while reducing the impact on the environment. In this context, determining a path for the robot that optimizes coverage while taking into account topography, robot characteristics, and operational requirements, is critical. In this paper, we present H-CCPP, a novel hybrid method that combines the comprehensive coverage benefits of our previous approach O-CCPP with the computational efficiency of the Fields2Cover algorithm. Besides optimizing coverage area, overlaps, and overall travel time, it significantly improves the computation process, and enhances the flexibility of trajectory generation. H-CCPP also considers terrain inclination to address soil erosion and energy consumption. In an effort to support this innovative approach, we have also created and made available a public data set that includes both 2D and 3D representations of 30 agricultural fields. This resource not only allows us to illustrate the effectiveness of our approach but also provides invaluable data for future research in complete coverage path planning (CCPP) for modern agriculture.

The underlying framework for controlling autonomous robots and complex automation applications is Operating Systems (OS) capable of scheduling perception-and-control tasks, as well as providing real-time data communication to other robotic peers and remote cloud computers. In this paper, we introduce CyberCortex.AI, a robotics OS designed to enable heterogeneous AI-based robotics and complex automation applications. CyberCortex.AI is a decentralized distributed OS which enables robots to talk to each other, as well as to High Performance Computers (HPC) in the cloud. Sensory and control data from the robots is streamed toward HPC systems with the purpose of training AI algorithms, which are afterwards deployed on the robots. Each functionality of a robot (e.g., sensory data acquisition, path planning, motion control, etc.) is executed within a so-called DataBlock of Filters shared through the internet, where each filter is computed either locally on the robot itself or remotely on a different robotic system. The data is stored and accessed via a so-called Temporal Addressable Memory (TAM), which acts as a gateway between each filter's input and output. CyberCortex.AI has two main components: (i) the CyberCortex.AI.inference system, which is a real-time implementation of the DataBlock running on the robots' embedded hardware, and (ii) the CyberCortex.AI.dojo, which runs on an HPC computer in the cloud, and it is used to design, train and deploy AI algorithms. We present a quantitative and qualitative performance analysis of the proposed approach using two collaborative robotics applications: (i) a forest fires prevention system based on an Unitree A1 legged robot and an Anafi Parrot 4K drone, as well as (ii) an autonomous driving system which uses CyberCortex.AI for collaborative perception and motion control.

Small carnivorous marine animals have developed agile movement abilities through long-term natural selection, resulting in excellent maneuverability and high swimming efficiency, making them ideal models for underwater robots. To meet the requirements for exploring narrow underwater zones, this paper designs an underwater robot inspired by mantis shrimp. By analyzing the body structure and swimming mode of the mantis shrimp, we designed a robot structure and hardware system and established a dynamic model for the coupled motion of multiple pleopods. A series of underwater experiments were conducted to verify the dynamic model and assess the performance of the prototype. The experimental results confirmed the accuracy of the dynamic model and demonstrated that the bionic mantis shrimp robot can perform multiangle turns and flexible velocity adjustments and exhibits good motion performance. This approach provides a novel solution for developing robots suitable for detecting complex underwater environments.

With the increasingly complex operating environment of mobile robots, the intelligent requirements of robots are getting higher and higher. Navigation technology is the core of mobile robot intelligent technology research, and path planning is an important function of mobile robot navigation. Dynamic window approach (DWA) is one of the most popular local path planning algorithms nowadays. However, there are also some problems. DWA algorithm is easy to fall into local optimal solution without the guidance of global path. The traditional solution is to use the key nodes of the global path as the temporary target points. However, the guiding ability of the temporary target points will be weakened in some cases, which still leads DWA to fall into local optimal solutions such as being trapped by a “C”-shaped obstacle or go around outside of a dense obstacle area. In a complex operating environment, if the local path deviates too far from the global path, serious consequences may be caused. Therefore, we proposed a trajectory similarity evaluation function based on dynamic time warping method to provide better guidance. The other problem is poor adaptability to complex environments due to fixed evaluation function weights. And, we designed a fuzzy controller to improve the adaptability of the DWA algorithm in complex environments. Experiment results show that the trajectory similarity evaluation function reduces algorithm execution time by 0.7% and mileage by 2.1%, the fuzzy controller reduces algorithm execution time by 10.8% and improves the average distance between the mobile robot and obstacles at the global path's danger points by 50%, and in simulated complex terrain environment, the finishing rate of experiments improves by 25%.

Point cloud registration is a vital task in three-dimensional (3D) perception, with several different applications in robotics. Recent advancements have introduced neural-based techniques that promise enhanced accuracy and robustness. In this paper, we thoroughly evaluate well-known neural-based point cloud registration methods using the Point Clouds Registration Benchmark, which was developed to cover a large variety of use cases. Our evaluation focuses on the performance of these techniques when applied to real-complex data, which presents a more challenging and realistic scenario than the simpler experiments typically conducted by the original authors. The results reveal considerable variability in performance across different techniques, highlighting the importance of assessing algorithms in realistic settings. Notably, 3DSmoothNet emerges as a standout solution, demonstrating good and consistent results across various data sets. Its efficacy, coupled with a relatively low graphics processing unit (GPU) memory footprint, makes it a promising choice for robotics applications, even if it is not yet suitable for real-time applications due to its execution time. Fully Convolutional Geometric Features also performs well, albeit with greater variability among data sets. PREDATOR and GeoTransformer are promising, but demand substantial GPU memory, when handling large point clouds from the Point Clouds Registration Benchmark. A notable finding concerns the performance of Fast Point Feature Histograms, which exhibit results comparable to the best approaches while demanding minimal computational resources. Overall, this comparative analysis provides valuable insights into the strengths and limitations of neural-based registration techniques, both in terms of the quality of the results and the computational resources required. This helps researchers to make informed decisions for robotics applications.

The coal industry has long been troubled by the imbalance between mining and tunneling, and between excavating and support. The main cause of this problem is the inability to perform excavating and permanent support operations in parallel. Additionally, the limited space near the tunneling face hampers the efficiency of permanent support. Temporary support is an effective method to ensure the stability of the surrounding rock and expand the space for parallel operations between excavating and permanent support activities. This work provides a brief overview of the current research status on temporary support, emphasizing that the key to achieving safe, efficient, and rapid excavation lies in the development of temporary support robots. To meet the development needs of temporary support robots, three key technologies are proposed: the construction of a coupling model between the robot and the surrounding rock, spatial layout optimization of the rapid advancement system, and adaptive control of the robot. This work details the methods and approaches for constructing the coupling model, the elements of system spatial layout optimization, and the methods and strategies for the robot's adaptive control. Our team successfully tested the shield robot system at Xiaobaodang Mining Company, verifying the feasibility of these key technologies.

Floor-tiling robotics are increasingly employed in on-site building constructions owing to their remarkable benefits on rising working efficiency and reducing labor costs. In this study, a fluid–structure interaction (FSI) model of robotic tiling was established for the first time, construction parameters and adhesive properties were modified, and their influences on the quality of robotic floor-tiling were systematically investigated by tracking the mechanical behaviors of tiles and adhesive during tiling and the interfacial defects after tiling. Results indicated that the established FSI model was feasible for assessing robotic tiling quality with a deviation of less than 2%. The adhesive extruded horizontally was evenly distributed in cylindrical strips. An increase in the number of extrusion pipes slightly improved the tiling quality. Compared with the leveling loads of compression and vertical vibration, shear vibration could effectively eliminate tile rebounding and enlarge the contact area of tile–adhesive by up to 135.85%. Moderate increases in the amplitude and frequency of shear vibration resulted in lower rebounding and larger contact areas. An appropriate increase of yield stress heightened tiling quality by keeping the extrusive appearance of the adhesive, increasing slightly tile rebounding and enlarging the contact area of tile–adhesive to 0.625 m². As yield stress was excessively high, tremendous elastic deformations of adhesive led to remarkable tile rebounding and small contact areas of 0.375 m².

To make the amphibious robot have a lot of functions while keeping the overall structure relatively simple, this paper proposes a multimodule bionic amphibious robot (MMBAR) inspired by the movement mode of jellyfish. The MMBAR consists of four modules, which are connected by snaps, and can be assembled quickly. The wing–leg structure suitable for swimming in the water is designed, which combines the legs and wings using a flexible hinge. Meanwhile, the integrated design principle is adopted to combine the wing–leg structure with the wheel structure to design a deformable wheel suitable for land movement. The overall structure of the MMBAR is simple, and the wing–legs can be deformed to perform a variety of functions, such as acting as a wheel for land movement, as a claw for grasping objects, and as a propulsion mechanism to power the MMBAR for swimming. Theoretical modeling and simulation analyses are conducted separately for the MMBAR on land and in water, which helps understand the movement characteristics of the MMBAR and to obtain more optimized movement parameters. In addition, we conducted experiments on the MMBAR, such as climbing slopes, climbing steps, walking on snow, swimming in water, grasping objects, and so forth, which confirm that the MMBAR possesses a strong ability to adapt to the environment. These research results add new content to the research of amphibious robots, which are expected to replace humans to fulfill more dangerous jobs.