Search Results (202)

Robust quadruped locomotion in real-world environments remains challenging because external disturbances, sensor noise, and model uncertainties are coupled with intermittent foot–ground contact. Reinforcement learning has shown strong capability in generating agile locomotion, but many existing methods handle unobserved disturbances through implicit latent representations or domain randomization. This paper presents a disturbance-aware locomotion framework that integrates state and disturbance estimation with learning-based control. The core component is a deep generalized-momentum-based Kalman filter, which combines generalized momentum disturbance modeling with adaptive covariance inference to estimate the base velocity and external disturbance force. These physically meaningful estimates are incorporated into the policy observation space, reducing the gap between privileged simulation states and deployable onboard observations. The framework was evaluated in a simulation and on a quadruped robot platform under disturbance and outdoor locomotion scenarios. Compared with the baseline and ablated variants, the proposed method reduced estimation and tracking errors, limited impact-induced torque peaks, and improved locomotion success rates under the evaluated conditions. The results suggest that explicit disturbance estimation can complement a latent adaptation for quadruped locomotion under impact-rich conditions. Full article

As robotics technology advances, quadruped robots have become capable of operating in complex environments with varying elevation, including ramps and level changes that are challenging for conventional wheeled platforms. While this terrain adaptability opens new opportunities for inspection, rescue, and exploration tasks, the repetitive impacts, frequent ground-contact transitions, and abrupt postural changes inherent to legged locomotion pose significant challenges for LiDAR odometry. High-frequency gait vibrations and abrupt attitude changes introduce intra-scan motion distortion that conventional single-twist deskewing cannot adequately suppress. In addition, sparse vertical geometric constraints in elevation-varying environments weaken Z-axis observability, allowing vertical drift to corrupt the horizontal pose estimate through Hessian coupling. To address these failure modes within a LiDAR-only framework, we propose a Piecewise-Constant Velocity deskewing scheme that partitions each scan into multiple temporal segments with safety clamping on vertical and attitude components, together with a two-stage ICP that decouples SE(3) optimization into horizontal (x, y, yaw) and vertical (z, roll, pitch) stages and applies observability-aware weighting in the vertical update. The proposed odometry front-end is evaluated on four real-world sequences collected with a Unitree Go2 quadruped robot equipped with a Velodyne VLP-16 LiDAR. Experimental results show consistently lower Absolute Pose Error (APE) than ICP, KISS-ICP, and F-LOAM across all sequences. Vertical drift suppression is most pronounced in the ramp-containing sequences, where baseline methods exhibit substantial Z-axis divergence. Full article

This paper proposes LA-PPO, an improved Proximal Policy Optimization algorithm for quadruped robot locomotion control on mixed terrain. To address partial observability, temporal dependence in contact states, and non-uniform importance of historical information in complex-terrain quadruped locomotion, LA-PPO integrates Long Short-Term Memory (LSTM) and Multi-Head Attention (MHA) within an Actor–Critic framework. The LSTM module models temporal dependencies in historical observations, while the MHA module adaptively emphasizes historical information most relevant to the current action decision. Based on IsaacGym, we construct a mixed-terrain environment consisting of flat regions, sloped regions, and random rough-terrain regions and conduct algorithmic comparisons, statistics over multiple random seeds, reward component ablation studies, and attention mechanism analyses for both walking and trotting gaits. Simulation results show that LA-PPO achieves the highest final reward and the longest mean episode length in both gaits. Compared with the PPO baseline, the final reward and mean episode length are improved by approximately 42.3% and 42.7%, respectively, in the walking task, and by approximately 39.8% and 25.7%, respectively, in the trotting task. Real-robot tests further show that the learned policy can perform walking and trotting on flat ground, sloped terrain, and random rough terrain, demonstrating preliminary sim-to-real transfer capability. Full article

Conventional quadruped robots are usually built with a rigid body, whereas quadrupedal mammals have flexible spines to perform agile behaviours on rough terrains. Applying a flexible spine to robots is a promising way to achieve dynamic and stable movement in extreme environments. In this paper, a novel bio-inspired spine constructed with a tensegrity structure is introduced. The prototype of the spine includes active and passive parts that can both be actively actuated and passively compliant. It has two joints with three degrees of freedom (DOF) each and can generate complex and multi-degree motions simultaneously. To control the spine with adjustable stiffness, a method based on vector closure and adjustment of pretension ratio is proposed. Several experiments are reported to illustrate the physical design of the spine and demonstrate the properties of the spine. The results demonstrate its capabilities of both active motion and passive compliance, which may improve adaptability in complex environments. Future work includes attachment of the spine to a quadruped robot to increase the overall workspace and generate rich motion skills. Full article

The demographic shift towards an aging population necessitates innovative solutions for care and mobility support. While commercial quadruped robots like the Unitree Go1 offer dynamic stability, their native following modes often lack the safety margins and predictability required, and they do not consistently follow the user, at times deviating and navigating independently. This paper presents a robust, vision-based, person-following algorithm designed to address these limitations. Utilizing a ZED 2 stereo camera and Robot Operating System (ROS), the system employs a finite state machine to ensure deterministic target tracking. A velocity control strategy partitions the robot’s motion into distinct stability, proportional, and braking zones based on depth data to ensure fluid interaction. The framework was validated on a Unitree Go1 quadruped platform in an outdoor environment involving 90-degree turns to evaluate tracking robustness. By operating in a headless mode, the system achieved a mean processing latency of $66.5\pm 4.3$ ms. Experimental results demonstrated consistent operational stability, 0.0% intrusion into the intimate safety zone, and effective velocity synchronization between $0.47$ and $0.54$ m/s. While this study establishes a robust technical baseline using healthy subjects, it serves as a preliminary development platform; further iterative testing with elderly users in clinical settings is required to move toward deployment. Beyond the evaluated trials, the framework maintained reliable functional performance across various care facility workshops, successfully following the target in all deployment scenarios. These findings establish a stable technical foundation for the future development of robotic walking partners. Full article

Benthic AUVs (underwater quadruped robots) merge the cruising efficiency of submersibles with the bottom-crawling stability of legged robots for unstructured deep-sea exploration. However, the deliberate separation of the center of gravity and buoyancy—essential for static stability—generates a significant righting moment. When climbing steep slopes, this moment resists hull alignment. If the slope exceeds the robot’s maximum hydrostatic pitch limit, conventional inverse kinematics algorithms fail: the hind legs lose ground contact and propulsion is lost. To overcome this, this paper proposes a framework integrating optimal force distribution, adaptive trajectory probing, and admittance control. An analytical multi-point moment balance model derives the terrain-adaptive pitch boundaries. A Quadratic Program (QP) then distributes contact forces, tasking front legs with stabilizing the righting moment while hind legs provide thrust. During the swing phase, adaptive Bezier sequences prevent anterior slope collisions and ensure posterior ground contact. Furthermore, a Cartesian admittance controller provides active compliance to manage the nonlinear friction of dynamic waterproof seals. Validated via a high-fidelity physics-based simulation model calibrated against physical pool trials, the robot achieved robust traversal of $15°$ and $33°$ steep slopes. Statistical robustness is substantiated via a 30-trial Monte Carlo study, where postural stability remained remarkably consistent with a mean Pitch RMSE of $2.88°$ across a $\pm 10\%$ parameter uncertainty envelope. Compared to traditional baseline algorithms, the proposed method successfully suppressed torque chattering by $54.1\%$ in the high-frequency band (2–$50\mathrm{Hz}$) and improved energetic efficiency by up to $43\%$ on steep gradients. These findings offer a validated control architecture for heavy-duty deep-sea platforms navigating complex benthic topographies. Full article

The integration of quadruped robotics with advanced sensing technologies offers a transformative approach to forest management, particularly for real-time measurement of tree Diameter at Breast Height (DBH). This paper introduces a novel methodology by deploying a quadruped robot equipped with GPS, LiDAR, and an aligned high-definition camera to patrol forest paths via a developed dynamic autonomous mission. Utilizing a YOLO-based model for trunk detection, the methodology retrieves precise DBH measurements and corresponding geotags, constructing a spatial database of DBH ground-truth data. This database serves as a real-time ground-truth lookup table to calibrate allometric equations used in drone-based crown detection missions, enhancing the accuracy of forest biophysical attribute estimations such as tree height, volume, and biomass. Experimental validation demonstrates high precision in DBH estimation (error < 5% in controlled tests), supporting automated, around-the-clock data collection for sustainable forest management in Mediterranean ecosystems. Full article

The spine enhances the flexibility of quadrupeds during locomotion. Inspired by this biological mechanism, this study incorporates an actuated spinal joint into a quadruped robot, enabling more natural motion and posture adjustment. To improve the motion stability of spinal robots in complex environments, a deep reinforcement learning framework that integrates a central pattern generator (CPG) with the twin delayed deterministic policy gradient (TD3) algorithm is proposed to optimize the gait motion of the spinal quadruped robot. First, the structure and parameters of the quadruped robot with a spinal joint are analyzed and a CPG coupling model incorporating spinal motion parameters is designed. Subsequently, a TD3–CPG algorithm framework based on a joint incremental strategy is proposed to optimize the robot’s gait, exploring optimal control strategies for terrain adaptation through spinal motion integration. Finally, experiments are conducted on various obstacle terrains to validate the proposed algorithm. Simulation and experiment results demonstrate the effectiveness of the algorithm in optimizing the gait of the spinal quadruped robot, showing significant improvements in walking stability, speed, and terrain adaptability across different terrains. Full article

Lunar quadruped robots face landing challenges including weak gravity, large mass variations, uncertain sloped terrain, and strict payload acceleration limits, requiring effective impact mitigation and rapid post-landing stabilization. This paper presents a novel end-to-end reinforcement learning-based landing controller with three key novelties: (i) a phase-structured yet implicitly encoded formulation that distinguishes contact preparation, energy dissipation, and stabilization without explicit phase switching; (ii) a terrain-agnostic state and control representation using equivalent support direction construction and contact-gated modulation to decouple normal–tangential dynamics; and (iii) an extremum oriented learning strategy that directly captures peak impact suppression and buffering sufficiency, addressing limitations of cumulative rewards in hybrid, peak-dominated tasks. A hybrid control model for lunar quadruped landing dynamics is established, incorporating variable mass, low impact, and full stroke as key constraints during training. Simulation and full-scale experimental prototypes are built to validate the controller. Simulation results demonstrate robust landing buffering and stability control under varying mass, landing velocity, and slope conditions, with favorable robustness against parameter variations. Experimental verification is conducted under diverse conditions including different masses (200 kg, 250 kg), vertical/horizontal landing velocities (0.8 m/s, 0.2 m/s), and slopes (0°, 8°). The deviation between simulation and experimental results does not exceed 30%, confirming the effectiveness and transferability of the proposed approach. Full article

This paper details the development of an autonomous robotic solution for the long-term surveillance of low-level radiation in nuclear facilities. Implementing such a system mitigates personnel health risks by minimizing radiation exposure and automating a mundane, repetitive task. To address the inherent challenges of deploying robots in highly unstructured environments, the core contribution of this work is a novel, error-tolerant behavioral architecture. Specifically, a custom behavior tree is designed to absorb execution imperfections and tolerate environmental uncertainties. This allows the robot to adapt and continue its mission rather than experiencing a hard failure. Bayesian optimization is utilized to perform adaptive mapping via a manipulator-equipped Spot quadruped robot, which features a Kromek Sigma50 gamma spectrometer attached to its end effector. Experiments were conducted in an obstacle-rich testbed using a Cesium-137 source. The results demonstrate the feasibility of the proposed system and its behavioral design approach, as the robot successfully performed adaptive mapping and correctly identified the location and approximate intensity of the radiation source. Full article

Agile parkour in unstructured environments poses significant challenges for quadruped robots, requiring both dynamic motion generation and terrain adaptability. Recent advances such as Adversarial Motion Priors (AMP) have shown promise in learning dynamic behaviors through motion imitation, but the resulting policies are typically specialized and struggle to generalize across varying terrains. However, existing AMP-based approaches largely lack explicit environmental awareness, leading to limited adaptability and revealing a clear gap in achieving general agile locomotion. To address this limitation, we propose a novel terrain-conditional AMP framework that extends adversarial motion priors by conditioning the discriminator on explicit terrain features, enabling the learning of terrain-aware motion representations adaptable to diverse environments. To improve practical applicability, we further leverage a vision-based policy distillation scheme, where a teacher policy with privileged terrain height information supervises a student policy operating only on forward-looking depth images. This enables agile, perception-driven locomotion in real time. To the best of our knowledge, this is the first work to integrate environmental information into adversarial motion priors and jointly learn a vision-based policy through policy distillation for agile quadruped locomotion. Experiments on terrains such as platforms, gaps, stairs, slopes, and debris show that the proposed method achieves more efficient training convergence and higher success rates compared to pure AMP-based and RL-based methods. These results highlight the effectiveness of the proposed framework and represent a step toward perception-driven agile locomotion for quadruped robots in complex environments. Full article

Bio-inspired quadrupedal robots exhibit superior adaptability and mobility in unstructured environments, making them suitable for complex task scenarios such as navigation, obstacle avoidance, and tracking in a variety of environments. Visual perception plays a critical role in enabling autonomous behavior, offering a cost-effective alternative to multi-sensor systems. This paper proposes an image segmentation-guided visual tracking framework to enhance both perception and motion control in quadruped robots. On the perception side, a cascaded convolutional neural network is introduced, integrating a global information guidance module to fuse low-level textures and high-level semantic features. This architecture effectively addresses limitations in single-scale feature extraction and improves segmentation accuracy under visually degraded conditions. On the control side, segmentation outputs are embedded into a biologically inspired central pattern generator (CPG), enabling coordinated generation of limb and spinal trajectories. This integration facilitates a closed-loop visual-motor system that adapts dynamically to environmental changes. Experimental evaluations on benchmark image segmentation datasets and robotic locomotion tasks demonstrate that the proposed framework achieves enhanced segmentation precision and motion flexibility, outperforming existing methods. The results highlight the effectiveness of vision-guided control strategies and their potential for deployment in real-time robotic navigation. Full article

As a typical biomimetic robotic system, quadruped robots replicate the flexible locomotion of quadruped mammals, outperforming wheeled robots in human-centered daily scenarios. To improve the social navigation adaptability of biomimetic quadruped robots in human–robot shared environments, this paper proposes a collision-aware orthogonal steering social force model (COSFM), an enhanced social force model that integrates collision prediction and social norms, inspired by human-like collision avoidance behaviors and social interaction rules. The model addresses key limitations of conventional social force models: delayed responses to dynamic pedestrians and inadequate consideration of pedestrians’ comfort zones. It introduces a time-to-collision prediction mechanism to mimic human predictive decision-making in dynamic social interactions, enhancing the robot’s anticipation of pedestrian motion intentions, and designs an orthogonal steering-based avoidance strategy for four typical human–robot interaction scenarios (head-on encounters, intersecting paths, active overtaking, passive yielding). This strategy replicates humans’ natural priority of lateral steering over abrupt deceleration or retreat, generating socially compliant trajectories aligned with human behavioral expectations. The proposed method is validated via simulation and real-world experiments on a Unitree Aliengo quadruped robot. Results show that the COSFM algorithm achieves a higher navigation success rate and better performance in path length, navigation time, and minimum human-robot distance than existing approaches, while its human-like lateral avoidance priority effectively preserves pedestrians’ psychological comfort zones, demonstrating robust social adaptability and great application potential for biomimetic legged robots. Full article

This paper presents an open-source ROS 2 navigation and sensor platform for quadruped robots, demonstrated on Boston Dynamics Spot in a laboratory environment. The platform integrates SLAM Toolbox for mapping and localisation, Navigation2 with MPPI and Smac Hybrid-A* for global path planning, and a frontier-based autonomous exploration module with practical handling of unreachable frontiers. The paper validates and verifies current, open-source algorithms deployed on off-the-shelf hardware. A greedy wavefront-based frontier selection method is presented that prioritizes Time-to-Closest-Viable-Frontier (TCVF) by terminating the search as soon as a feasible frontier is identified. On a real robot dataset replayed across five goal scenarios, the method reduces median selection latency from $94.31$ ms to $51.08$ ms (95th percentile: $109.54$ ms to $56.99$ ms), corresponding to a $1.85$-times improvement in compute time compared to a standard implementation. The system also employs Zenoh middleware and Foxglove for remote monitoring and control, enabling flexible, high-bandwidth operation. The platform, including configuration files and launch scripts, is released openly to support future research and deployment on quadruped robots. Full article

This paper studies a MuJoCo-based locomotion framework that couples an adaptive-frequency central pattern generator (AFCO-CPG) with single rigid-body dynamics model predictive control (MPC) for the RENS Q1 quadruped with elastic parallel knee joints. AFCO-CPG combines multi-scale phase coordination, saturated phase correction, and load-gated feedback, while MPC supplies feasible ground-reaction forces and returns load cues to the timing layer. In MuJoCo, the controller achieves stable diagonal-trot speed tracking from 0.4 to 1.2 m/s and recovers from short external pushes. A matched elastic-versus-rigid timing sweep shows a favorable flat-ground parameter band around $\omega =1.8$ Hz, with a best-case cost-of-transport reduction of 12.83% for the elastic model under identical controller gains. A flat-to-slope ascent case further verifies that AFCO timing is modulated when load conditions change. Ablation across nine controller variants shows that multi-scale coordination is the dominant component, causing a 135% increase in phase error and a 536% increase in recovery time when removed. A reduced-order early/late-contact benchmark further confirms faster re-locking than diagonal-only and minimal variants. The results support the value of combining neural timing, predictive force optimization, and compliant-leg feedback in high-fidelity simulation, while hardware validation remains future work. Full article