Journal of Field Robotics最新文献

Large-scale semantic mapping is crucial for outdoor autonomous agents to perform high-level tasks such as planning and navigation. In this paper, we propose a novel method for large-scale 3D semantic reconstruction through implicit representations from posed LiDAR measurements alone. We first construct a neural fields via the octree and latent features, then decode implicit features into signed distance value and semantic information through shallow Multilayer Perceptrons (MLPs). We leverage radial window self-attention networks to predict the semantic labels of point clouds. We then jointly optimize the feature embedding and MLP parameters with a self-supervision paradigm for point-cloud geometry and a pseudo-supervision paradigm for semantic and panoptic labels. Subsequently, geometric structures and semantic categories for novel points in the unseen area are regressed, and the marching cubes method is exploited to subdivide and visualize scenes in the inferring stage, the labeled mesh is produced correspondingly. Experiments on two real-world datasets, SemanticKITTI and SemanticPOSS, demonstrate the superior segmentation efficiency and mapping effectiveness of our framework compared to existing 3D semantic LiDAR mapping methods.

Heavy-duty robots such as hexapod robots exhibit tremendous potential with their high payload ability and terrain adaptability. This paper introduces a control approach for heavy-duty hexapod robot that traverses uneven and transition stage from a flat surface to a sloped terrain without prior geometric knowledge of the environment. Using force sensors on the robot feet and position sensors at its joints, the transition from a flat surface to a sloped terrain is detected. Constraint models based on terrain geometric information and joint limit positions are presented. The models are derived to estimate the unknown slope gradient based on the landing feet positions. The constraints for different terrains have also been investigated. An optimal force allocation method under contact constraints is developed to minimize the linear and quadratic costs in the commands and constrained contact forces. Extensive experiments have been conducted, and the results have demonstrated the effectiveness of the proposed control approach.

The flight efficiency of bionic flapping-wing robots is largely influenced by the flapping mechanism. This paper presents the development of a large wingspan flapping-wing flying robot that utilizes a rigid-flexible coupling flapping mechanism to reduce energy consumption, improve flight efficiency, and enhance wind resistance. The rigid-flexible coupling flapping mechanism consists of a DC motor with a gearset and a flexible flapping mechanism based on torsion spring. This mechanism combines the high torque driving capability of a rigid mechanism with the energy storage capacity of a flexible mechanism. Synchronizing the passive deformation of the torsion spring with the periodic acceleration and deceleration of the wings during the flapping cycle enables the storage, transfer, and release of kinetic energy and torsional spring elastic strain energy. Simulation studies show that the proposed design effectively reduces the peak power required. A prototype with a wingspan of 1.8 meters and a mass of 1.0 kilograms was developed. Compared to a rigid mechanism, the proposed design significantly reduces the electrical power consumption, especially when achieving flapping frequencies exceeding 1.5 Hz, as validated through wind tunnel experiments. At a flapping frequency of 2.2 Hz, the maximum reduction in electrical power consumption reaches 14.8%. The flexible elements also increased the downstroke ratio, consequently enhancing thrust and propulsion efficiency. Outdoor flight experiments have demonstrated that the prototype propelled by the rigid-flexible coupling drive mechanism consumes only 81.56 W during cruising, which is 8.78 W (9.72%) less than the rigidly driven prototype.

Ice cover hinders the deployment of acoustic arrays or satellites, making it challenging to locate underwater vehicles in the polar regions. Odometry based on acoustic or optical sensors has the potential to provide precise navigation results for vehicles. However, the robustness of single-sensor-based odometry is significantly compromised by the smooth and featureless nature of the ice-covered bottom, which lacks distinct points, lines, or surfaces. This paper presents a filter-based multi-sensor fusion underwater ice odometry system, integrating measurements from a stereo camera, a forward-looking sonar (FLS), a low-cost inertial measurement unit (IMU), and a depth gauge (DG). To balance real-time performance and accuracy, the Multi-State Constraint Kalman Filter (MSCKF) algorithm is employed to fuse inertial and stereo measurements for vehicle state estimation. Additionally, FLS images and relative depth measurements from the DG are utilized to enhance the accuracy and robustness of the state estimation. Specifically, a Fourier-based algorithm is proposed to estimate linear speed using global information from FLS images. A compensation factor is introduced to address the impact of FLS elevation angle on estimation accuracy. Furthermore, a residual-based adaptive method and forgetting factor are implemented to adjust the measurement noise covariance of linear speed and relative depth, improving dynamic state estimation precision. The proposed system was evaluated through tank and field experiments using a remotely operated vehicle (ROV). Results show it provides more accurate and robust under-ice navigation compared to state-of-the-art algorithms.

Despite the increasing prevalence of robots in daily life, their navigation capabilities are still limited to environments with prior knowledge, such as a global map. To fully unlock the potential of robots, it is crucial to enable them to navigate in large-scale unknown and changing unstructured scenarios. This requires the robot to construct an accurate static map in real-time as it explores, while filtering out moving objects to ensure mapping accuracy and, if possible, achieving high-quality pedestrian tracking and collision avoidance. While existing methods can achieve individual goals of spatial mapping or dynamic object detection and tracking, there has been limited research on effectively integrating these two tasks, which are actually coupled and reciprocal. In this study, we propose a solution called S²MAT (simultaneous and self-reinforced mapping and tracking) that integrates a front-end dynamic object detection and tracking module with a back-end static mapping module. S²MAT leverages the close and reciprocal interplay between these two modules to efficiently and effectively solve the open problem of simultaneous tracking and mapping in highly dynamic scenarios. The proposed method is primarily designed for use with 3D LiDAR and offers a solution for real-time navigation in large-scale, unknown dynamic scenarios with a low computational cost, making it feasible for deployment on onboard computers equipped with only a single CPU. We conducted long-range experiments in real-world urban scenarios spanning over 7 km, which included challenging obstacles like pedestrians and other traffic agents. The successful navigation provides a comprehensive test of S²MAT's robustness, scalability, efficiency, quality, and its ability to benefit autonomous robots in wild scenarios without pre-built maps.

The efficacy of field exploration robots is contingent upon their capacity and proficiency in traversing unstructured and complex terrains. In the context of beach terrain, substantial variations in the mechanical properties of the substrate can be triggered by fluctuations in moisture levels within granular materials. This poses a substantial challenge to the effectiveness of actuators designed for amphibious crawling, which depend exclusively on plowing force after penetration or surface friction to adapt to such environments. In this study, we propose a cost-effective, adaptable wheel-foot actuator that is designed based on the mechanical properties of granular media. The configuration consists of a circular wheel as the primary structural element and a regular arrangement of flexible feet, which collectively provide a combined traction force of penetration and friction. A total of 3840 experiments were conducted under controlled conditions on both the flexible wheel-foot scheme and the grousered wheel scheme. The outcomes of these experiments enabled us to ascertain and validate the penetration difficulty tendency of wet granules with varying moisture contents. Secondly, a comparison of the transportation cost, normalized antislip coefficient, standard deviation of vertical displacement, energy consumption, and other parameters of 40 schemes under the same working conditions allowed us to determine the structural parameters, such as the number and stiffness of flexible feet. This comparison also demonstrated that the proposed scheme has significant advantages in suppressing slip and reducing energy consumption. In field tests conducted in an amphibious environment, the proposed scheme demonstrated a statistically significant reduction in slip rate (10.4%) and energy consumption (22.8%) compared to other tested schemes. Subsequent evaluations were conducted through tests of water-land transition capability, extrication ability, and the ability to navigate through harsh ground conditions. These evaluations confirmed the effectiveness of the proposed scheme.

Simultaneous localization and mapping (SLAM) is at the core of robotics automation, relying on sensors such as Laser Detection and Ranging (LiDAR) and cameras to digitally construct a robot's environment and determine its position within. LiDAR-based SLAM outperforms visual-SLAM, especially in low visibility and challenging lighting conditions. However, these systems still face challenges like scene degradation when dealing with feature-deficient degenerate environments such as long corridors or tunnels. Traditional LiDAR SLAM algorithms primarily focus on the extraction of geometric features from the scene, with less utilization of visual information, for example, LiDAR-generated reflectivity (also commonly referred to as intensity image) and depth imagery. In this study, we explore the potential of fusing both geometric and LiDAR-generated image features into the SLAM system in various forms, aiming to enhance the system's adaptability in diverse environments and its robustness against environment degeneracy. We propose a new multifeature-modality SLAM designed for robust real-time localization and mapping in challenging environments. Our method enhances and extracts visual features from LiDAR-generated images, which are then fused with geometric features through a holistic residual function for pose optimization. We also integrate a deep learning-based object removal algorithm to reduce sensitivity to moving objects and sensor noise. This article conducts an in-depth comparison of the proposed algorithm with several leading technologies in terms of scan matching accuracy, robustness, odometry, and mapping. The experimental results vividly showcase the superiority of our method in achieving high scan matching success rates and strong resilience against random outliers and Gaussian noise across various challenging scenarios, compared to the existing LiDAR SLAM methods that rely solely on geometric features. Extensive field experiments conducted on publicly available data sets, along with independently developed backpack-based and robotic platforms, validated the robustness and accuracy of the proposed approach in both indoor and outdoor environments. In 3D mapping, we quantified the precision of 3D points by comparing point clouds collected by high-precision Mobile Laser Scanning (MLS) and Terrestrial Laser Scanning (TLS). Our method outperforms in terms of absolute pose errors (APE) and point cloud matching quality. Based on the fitted Weibull distribution, the root mean square error (RMS) of point-to-plane distances improved by 20%. Additionally, ablation tests revealed the efficacy of different components within our system.

In cotton picking, there is a need for a robotic arm that can pick bolls efficiently with minimal loss and damage to immature bolls. This study addresses that gap by developing a hydraulic arm to work effectively in unstructured cotton fields. A three-degrees-of-freedom hydraulic arm was developed with vertical, radial, and rotational movements, equipped with a roller-type end-effector and vacuum suction system to pick cotton bolls from the plants and collect them into a storage tank. The performance of the hydraulic arm was evaluated in the field under varying conditions of plant-to-plant spacing (250, 350, and 450 mm) and at suction pressures (25, 50, and 75 mmHg). Key performance parameters such as picking efficiency, picking capacity, damage of green bolls, losses of cotton bolls, and trash content were evaluated. The results indicated that the maximum picking efficiency (92.79%) and capacity (567 bolls picked/h) were achieved at 450 mm plant-to-plant spacing and 75 mmHg suction pressure, with minimal damage of green bolls (126 bolls/ha) and losses of cotton bolls (3.79%). The lowest trash content (2.06%) was observed at 450 mm spacing and 25 mmHg suction pressure. Finally, the hydraulic arm achieved precise and accurate cotton picking in complex cotton fields, showing the best performance at 450 mm plant-to-plant spacing and 75 mmHg suction pressure.