Search Results (33)

As autonomous driving technology expands from structured urban roads to unstructured outdoor environments, precise understanding of complex terrain has become a critical requirement for ensuring safe vehicle navigation. However, outdoor environments are characterized by high dynamics, drastic illumination variations, ambiguous category boundaries, and prohibitive annotation costs, making traditional supervised learning methods that rely on large amounts of pixel-level annotations difficult to generalize. In this paper, we propose a novel dual-path bidirectional interactive encoder, termed LapDINO, that effectively combines the strong semantic generalization capability of the self-supervised foundation model DINOv3 with the multi-scale frequency analysis capacity of the Laplacian pyramid. Specifically, we leverage DINOv3 to extract global semantic features as a “semantic map”, while simultaneously obtaining multi-scale high-frequency details through Laplacian pyramid decomposition as “structural contours”. Building upon this, we design a bidirectional cross-attention fusion mechanism that enables dynamic interaction and mutual refinement between semantic information and geometric details. Furthermore, we introduce a multi-branch attention enhancement module that extracts pyramid features from three complementary perspectives. To address domain shift, we design lightweight visual adapters that enable efficient fine-tuning of the frozen DINOv3 backbone. Finally, we construct two off-road terrain segmentation datasets, VOTD and VOCD, to facilitate research in this domain. Experimental results demonstrate that the proposed method achieves state-of-the-art performance, striking an optimal balance between accuracy and computational efficiency, thereby providing a robust and efficient engineering solution for terrain perception in off-road environments. Full article

Autonomous navigation in unstructured off-road environments presents fundamental challenges due to terrain heterogeneity, the absence of structured road markings, and the necessity for real-time traversability reasoning from raw sensory observations. We present TerrainFormer, a hierarchical framework that integrates a world model for terrain dynamics prediction with a temporal decision transformer for action selection. Our methodology employs a two-phase training paradigm: (1) self-supervised world model pretraining on LiDAR point clouds to learn terrain representations encompassing traversability, elevation, and semantic segmentation; (2) behavioral cloning of the decision transformer conditioned on frozen world model features with temporally derived goal directions. The world model processes raw 3D LiDAR point clouds through a PointPillars encoder for real-time bird’s-eye-view (BEV) projection, followed by a Vision Transformer backbone that produces latent terrain representations. A principal contribution is our cross-dataset generalization paradigm: the world model is trained on separate datasets while the decision transformer is trained on separate sequences, ensuring zero data overlap between training phases. We introduce automatic goal direction computation from vehicle pose trajectories, enabling the model to learn directionally conditioned navigation policies. To address the class imbalance inherent in off-road driving data, we employ focal loss with inverse-frequency class weighting and action-chunk supervision. Experimental evaluation on the RELLIS-3D dataset achieves 87.31% test accuracy with 0.7948 macro F1 across all 12 action classes. The world model’s predicted future frames produce only a 0.79% accuracy drop versus ground-truth observations, with 98.82% action agreement, demonstrating effective cross-dataset generalization for real-time off-road navigation. Full article

Autonomous navigation on unknown uneven terrain remains a critical challenge for unmanned ground vehicle (UGV) deployed in unstructured environments such as disaster relief, wilderness exploration, and off-road logistics. Existing motion planning methods for such environments suffer from three key limitations: under-utilization of the solution space due to discretized terrain assessment, difficulty in transforming complex terrain safety constraints into optimization-compatible forms, and the inherent trade-off between environmental modeling accuracy and real-time performance. This paper presents a hierarchical motion planning framework that enables safe and fast navigation of UGV on unknown uneven terrain. We first construct a traversability map based on terrain slope, roughness, and sparsity extracted from ground point cloud clusters. Non-traversable points are then transformed via spherical inversion and inverse mapping to generate terrain safety corridors composed of a series of convex polygons. The geometric containment relationship between the vehicle’s convex hull and the corridor is reformulated as continuously differentiable Control Barrier Function (CBF) constraints to ensure driving safety. The front-end employs a kinodynamic Hybrid A* algorithm with a traversability-aware node pruning strategy, while the back-end trajectory optimization embeds the CBF constraints as hard constraints within the optimization loop to guarantee forward invariance of the safety set under the linearized dynamics. The proposed framework achieves full-shape collision avoidance without sacrificing the solution space, while maintaining real-time performance for autonomous navigation on complex terrain. Full article

Autonomous systems are expanding rapidly, driving a demand for robust perception technologies capable of navigating challenging, unstructured environments. While urban autonomy has made significant progress, off-road environments pose unique challenges, including dynamic terrain and limited communication infrastructure. This research addresses these challenges by introducing a novel three-part solution for off-road autonomous vehicles. First, we present a large-scale off-road dataset curated to capture the visual complexity and variability of unstructured environments, providing a realistic training ground that supports improved model generalization. Second, we propose a Confusion-Aware Loss (CAL) that dynamically penalizes systematic misclassifications based on class-level confusion statistics. When combined with cross-entropy, CAL improves segmentation mean Intersection over Union (mIoU) on the off-road test set from 68.66% to 70.06% and achieves cross-domain gains of up to ~0.49% mIoU on the Cityscapes dataset. Third, leveraging semantic segmentation as an intermediate representation, we introduce a spatial overlay video encoding scheme that preserves high-fidelity RGB information in semantically critical regions while compressing non-essential background regions. Experimental results demonstrate Peak Signal-to-Noise Ratio (PSNR) improvements of up to +5 dB and Video Multi-Method Assessment Fusion (VMAF) gains of up to +40 points under lossy compression, enabling efficient and reliable off-road autonomous operation. This integrated approach provides a robust framework for real-time remote operation in bandwidth-constrained environments. Full article

Autonomous mobile robots require robust traversability perception to navigate safely in diverse outdoor environments. However, traditional deep learning approaches are data-hungry, requiring large-scale manual annotations, and struggle to adapt quickly to unseen environments. This paper introduces TARTS (Training-free Adaptive Reference-guided Traversability Segmentation), a novel framework combining one-shot prototype initialization with trajectory-guided online adaptation for terrain segmentation. Using a single reference image of desired traversable terrain, TARTS establishes an initial prototype from pre-trained DINO Vision Transformer (ViT) features. The system performs segmentation through superpixel-based feature aggregation and valley-emphasis Otsu thresholding while continuously refining the prototype via Exponential Moving Average (EMA) updates driven by automated footprint supervision from the robot’s traversed trajectory. Extensive experiments on our introduced Reference-guided Traversability Segmentation Dataset (RTSD) and the challenging Off-Road Freespace Detection (ORFD) benchmark demonstrate strong performance, achieving 94.5% IoU on RTSD and 94.1% IoU on ORFD, outperforming state-of-the-art supervised methods that require multi-modal inputs and dedicated training. The framework maintains efficient performance (17–24 FPS) on embedded platforms, enabling practical deployment with only a reference image as initialization. Full article

In contrast to structured urban settings, road networks in post-disaster or unstructured wildland environments are often incomplete or compromised. Navigation in these contexts requires navigating complex terrains and mitigating potential hazards that impede unmanned ground vehicles (UGVs). While high-mobility off-road vehicles are specifically designed to traverse challenging features like ditches and steep slopes, traditional path planning algorithms often fail to exploit these capabilities. These algorithms typically suffer from a binary focus, either relying strictly on road networks or ignoring them altogether, thereby neglecting the synergy between infrastructure and vehicle mobility. This chapter introduces a global path planning method based on traversability analysis and terrain matching to bridge this gap. The methodology incorporates a grid-based traversability evaluation, a road network expansion algorithm for densifying critical segments, and a unified planning strategy. By correlating terrain characteristics with vehicle mobility limits and optimizing the road network density, the proposed framework achieves an integrated on-road and off-road planning solution that maximizes the operational efficiency of high-mobility vehicles in degraded environments. Full article

Uneven terrains and complex obstacles in off-road environments present significant challenges to the stability and safety of vehicle path planning. This paper presents a hierarchical DRL-DWA path planning framework for unmanned ground vehicles (UGVs). At the global level, an energy-aware D* Lite algorithm generates cost-efficient waypoints considering both distance and energy consumption. At the local level, a deep reinforcement learning-enhanced DWA controller adaptively adjusts the weighting factors of evaluation functions in real time to ensure dynamic feasibility on rough terrain. The parameter selection is formulated as a Markov decision process (MDP), where a novel reward function based on elevation maps, vehicle pose, goal, and obstacle information guides the optimization for off-road navigation. Furthermore, the single UGV framework is extended to a multi-UGV system, where formation control is achieved through the leader–follower strategy. To evaluate the performance of our algorithm, we conduct experiments in 3D simulation environments featuring various terrains and obstacles. The results indicate that the proposed approach outperforms existing path planning techniques, showing a higher success rate and a lower average elevation gradient in uneven terrains. Full article

Autonomous exploration of autonomous vehicles in off-road environments remains challenging due to the adverse impact on exploration efficiency and safety caused by uneven terrain. In this paper, we propose a path planning framework for autonomous exploration to obtain feasible and smooth paths for autonomous vehicles in 3D off-road environments. In our framework, we design a target selection strategy based on 3D terrain traversability analysis, and the traversability is evaluated by integrating vehicle dynamics with geometric indicators of the terrain. This strategy detects the frontiers within 3D environments and utilizes the traversability cost of frontiers as the pivotal weight within the clustering process, ensuring the accessibility of candidate points. Additionally, we introduced a more precise approach to evaluate navigation costs in off-road terrain. To obtain a smooth local path, we generate a cluster of local paths based on the global path and evaluate the optimal local path through the traversability and smoothness of the path. The method is validated in simulations and real-world environments based on representative off-road scenarios. The results demonstrate that our method reduces the exploration time by up to 36.52% and ensures the safety of the vehicle while exploring unknown 3D off-road terrain compared with state-of-the-art methods. Full article

The motion control of a tracked mobile robot remains an important capability for autonomous navigation. Kinematic path-tracking algorithms are commonly used in mobile robotics due to their ease of implementation and real-time computational cost advantage. This paper integrates an extended Kalman filter (EKF) into a common kinematic controller for path-tracking performance improvement. The extended Kalman filter estimates the instantaneous center of rotation (ICR) of tracks using the sensor readings of GPS and IMU. These ICR estimations are then given as input to the motion control algorithm to generate the track velocity demands. The platform to be controlled is a heavyweight off-road tracked vehicle, which necessitates the investigation of slip values. A high-fidelity simulation model, which is verified with field tests, is used as the plant in the path-tracking simulations. The performance of the filter and the algorithm is also demonstrated in field tests on a stabilized road. The field results show that the proposed estimation increases the path-tracking accuracy significantly (about 44%) compared to the classical pure pursuit. Full article

Off-road path planning involves navigating vehicles through areas lacking established road networks, which is critical for emergency response in disaster events, but is limited by the complex geographical environments in natural conditions. How to model the vehicle’s off-road mobility effectively and represent environments is critical for efficient path planning in off-road environments. This paper proposed an improved A* path planning algorithm based on a Delaunay triangular NavMesh model with off-road environment representation. Firstly, a land cover off-road mobility model is constructed to determine the navigable regions by quantifying the mobility of different geographical factors. This model maps passable areas by considering factors such as slope, elevation, and vegetation density and utilizes morphological operations to minimize mapping noise. Secondly, a Delaunay triangular NavMesh model is established to represent off-road environments. This mesh leverages Delaunay triangulation’s empty circle and maximum-minimum angle properties, which accurately represent irregular obstacles without compromising computational efficiency. Finally, an improved A* path planning algorithm is developed to find the optimal off-road mobility path from a start point to an end point, and identify a path triangle chain with which to calculate the shortest path. The improved road-off path planning A* algorithm proposed in this paper, based on the Delaunay triangulation navigation mesh, uses the Euclidean distance between the midpoint of the input edge and the midpoint of the output edge as the cost function $g\left(n\right)$, and the Euclidean distance between the centroids of the current triangle and the goal as the heuristic function $h\left(n\right)$. Considering that the improved road-off path planning A* algorithm could identify a chain of path triangles for calculating the shortest path, the funnel algorithm was then introduced to transform the path planning problem into a dynamic geometric problem, iteratively approximating the optimal path by maintaining an evolving funnel region, obtaining a shortest path closer to the Euclidean shortest path. Research results indicate that the proposed algorithms yield optimal path-planning results in terms of both time and distance. The navigation mesh-based path planning algorithm saves 5~20% of path length than hexagonal and 8-directional grid algorithms used widely in previous research by using only 1~60% of the original data loading. In general, the path planning algorithm is based on a national-level navigation mesh model, validated at the national scale through four cases representing typical natural and social landscapes in China. Although the algorithms are currently constrained by the limited data accessibility reflecting real-time transportation status, these findings highlight the generalizability and efficiency of the proposed off-road path-planning algorithm, which is useful for path-planning solutions for emergency operations, wilderness adventures, and mineral exploration. Full article

This paper proposes a novel navigation approach for Unmanned Tracked Vehicles (UTVs) using prior-based ensemble deep reinforcement learning, which fuses the policy of the ensemble Deep Reinforcement Learning (DRL) and Dynamic Window Approach (DWA) to enhance both exploration efficiency and deployment safety in unstructured off-road environments. First, by integrating kinematic analysis, we introduce a novel state and an action space that account for rugged terrain features and track–ground interactions. Local elevation information and vehicle pose changes over consecutive time steps are used as inputs to the DRL model, enabling the UTVs to implicitly learn policies for safe navigation in complex terrains while minimizing the impact of slipping disturbances. Then, we introduce an ensemble Soft Actor–Critic (SAC) learning framework, which introduces the DWA as a behavioral prior, referred to as the SAC-based Hybrid Policy (SAC-HP). Ensemble SAC uses multiple policy networks to effectively reduce the variance of DRL outputs. We combine the DRL actions with the DWA method by reconstructing the hybrid Gaussian distribution of both. Experimental results indicate that the proposed SAC-HP converges faster than traditional SAC models, which enables efficient large-scale navigation tasks. Additionally, a penalty term in the reward function about energy optimization is proposed to reduce velocity oscillations, ensuring fast convergence and smooth robot movement. Scenarios with obstacles and rugged terrain have been considered to prove the SAC-HP’s efficiency, robustness, and smoothness when compared with the state of the art. Full article

Hyperspectral imaging (HSI) has evolved from its origins in space missions to become a promising sensing technology for mobile ground robots, offering unique capabilities in material identification and scene understanding. This review examines the integration and applications of HSI systems in ground-based mobile platforms, with emphasis on outdoor implementations. The analysis covers recent developments in two main application domains: autonomous navigation and inspection tasks. In navigation, the review explores HSI applications in Advanced Driver Assistance Systems (ADAS) and off-road scenarios, examining how spectral information enhances environmental perception and decision making. For inspection applications, the investigation covers HSI deployment in search and rescue operations, mining exploration, and infrastructure monitoring. The review addresses key technical aspects including sensor types, acquisition modes, and platform integration challenges, particularly focusing on environmental factors affecting outdoor HSI deployment. Additionally, it analyzes available datasets and annotation approaches, highlighting their significance for developing robust classification algorithms. While recent advances in sensor design and processing capabilities have expanded HSI applications, challenges remain in real-time processing, environmental robustness, and system cost. The review concludes with a discussion of future research directions and opportunities for advancing HSI technology in mobile robotics applications. Full article

Navigating uneven, unstructured terrain with dynamic obstacles remains a challenge for autonomous mobile robots. This article introduces Dynamic Unstructured Environment (DUnE) for evaluating the performance of off-road navigation systems in simulation. DUnE is a versatile software framework that implements the Gymnasium reinforcement learning (RL) interface for ROS 2, incorporating unstructured Gazebo simulation environments and dynamic obstacle integration to advance off-road navigation research. The testbed automates key performance metric logging and provides semi-automated trajectory generation for dynamic obstacles including simulated human actors. It supports multiple robot platforms and five distinct unstructured environments, ranging from forests to rocky terrains. A baseline reinforcement learning agent demonstrates the framework’s effectiveness by performing pointgoal navigation with obstacle avoidance across various terrains. By providing an RL interface, dynamic obstacle integration, specialized navigation tasks, and comprehensive metric tracking, DUnE addresses significant gaps in existing simulation tools. Full article

LiDAR-based place recognition (LPR) is crucial for the navigation and localization of autonomous vehicles and mobile robots in large-scale outdoor environments and plays a critical role in loop closure detection for simultaneous localization and mapping (SLAM). Existing LPR methods, which utilize 2D bird’s-eye view (BEV) projections of 3D point clouds, achieve competitive performance in efficiency and recognition accuracy. However, these methods often struggle with capturing global contextual information and maintaining robustness to viewpoint variations. To address these challenges, we propose R2SCAT-LPR, a novel, transformer-based model that leverages self-attention and cross-attention mechanisms to extract rotation-robust place feature descriptors from BEV images. R2SCAT-LPR consists of three core modules: (1) R2MPFE, which employs weight-shared cascaded multi-head self-attention (MHSA) to extract multi-level spatial contextual patch features from both the original BEV image and its randomly rotated counterpart; (2) DSCA, which integrates dual-branch self-attention and multi-head cross-attention (MHCA) to capture intrinsic correspondences between multi-level patch features before and after rotation, enhancing the extraction of rotation-robust local features; and (3) a combined NetVLAD module, which aggregates patch features from both the original feature space and the rotated interaction space into a compact and viewpoint-robust global descriptor. Extensive experiments conducted on the KITTI and NCLT datasets validate the effectiveness of the proposed model, demonstrating its robustness to rotation variations and its generalization ability across diverse scenes and LiDAR sensors types. Furthermore, we evaluate the generalization performance and computational efficiency of R2SCAT-LPR on our self-constructed OffRoad-LPR dataset for off-road autonomous driving, verifying its deployability on resource-constrained platforms. Full article

Topography estimation is essential for autonomous off-road navigation. Common methods rely on point cloud data from, e.g., Light Detection and Ranging sensors (LIDARs) and stereo cameras. Stereo cameras produce dense point clouds with larger coverage but lower accuracy. LIDARs, on the other hand, have higher accuracy and longer range but much less coverage. LIDARs are also more expensive. The research question examines whether incorporating LIDARs can significantly improve stereo camera accuracy. Current sensor fusion methods use LIDARs’ raw measurements directly; thus, the improvement in estimation accuracy is limited to only LIDAR-scanned locations The main contribution of our new method is to construct a reference ground plane through the interpolation of LIDAR data so that the interpolated maps have similar coverage as the stereo camera’s point cloud. The interpolated maps are fused with the stereo camera point cloud via Kalman filters to improve a larger section of the topography map. The method is tested in three environments: controlled indoor, semi-controlled outdoor, and unstructured terrain. Compared to the existing method without LIDAR interpolation, the proposed approach reduces average error by 40% in the controlled environment and 67% in the semi-controlled environment, while maintaining large coverage. The unstructured environment evaluation confirms its corrective impact. Full article