Trajectory planning technology is one of the key technologies in the field of autonomous driving. However, the current planning algorithms cannot meet the optimal trajectory requirements to a certain extent. A novel algorithm is proposed to optimize the A*-artificial potential field method for generating optimal trajectories. To address the issue of path non-optimality, improvements are made to the node expansion in the traditional A* algorithm. Additionally,the traditional four-connected search strategy is enhanced with a new hybrid search strategy. Nodes are pruned to reduce path length. To handle multi-turns in the A* algorithm, the trajectory is smoothed using a third-order Bezier curve, ensuring that the curvature of the trajectory is continuous. To address the issue of invalid repulsive force existing in the Artificial Potential Field (APF), the virtual ellipse theory is proposed. This theory aims to eliminate the impact of invalid repulsive force within certain ranges. At the same time, constraints such as road boundary repulsive force and virtual lane line gravitational force are incorporated to ensure safe vehicle travel. Finally, the optimized A*-APF algorithm is proposed to introduce an artificial potential energy term in the heuristic function of the A* algorithm to optimize trajectory generation. The algorithm is also verified in three scenarios on a real vehicle: vehicle and pedestrian avoidance experiment, parallel obstacle avoidance experiment, and staggered stopping experiment. The effectiveness of the algorithm is verified through the analysis of four parameters, namely trajectory, speed, heading angle, and steering wheel angle.

3D object-detection based on LiDAR point clouds can help driverless vehicles detect obstacles. However, the existing point-cloud-based object-detection methods are generally ineffective in detecting small objects such as pedestrians and cyclists. Therefore, a small-object-detection algorithm based on clustering is proposed. Firstly, a new segmented ground-point clouds segmentation algorithm is proposed, which filters out the object point clouds according to the heuristic rules and realizes the ground segmentation by multi-region plane-fitting. Then, the small-object point cloud is clustered using an improved DBSCAN clustering algorithm. The K-means++ algorithm for pre-clustering is used, the neighborhood radius is adaptively adjusted according to the distance, and the core point search method of the original algorithm is improved. Finally, the detection of small objects is completed using the directional wraparound box model. After extensive experiments, it was shown that the precision and recall of our proposed ground-segmentation algorithm reached 91.86% and 92.70%, respectively, and the improved DBSCAN clustering algorithm improved the recall of pedestrians and cyclists by 15.89% and 9.50%, respectively. In addition, visualization experiments confirmed that our proposed small-object-detection algorithm based on the point-cloud clustering method can realize the accurate detection of small objects.

With the rapid progress of intelligent vehicle technology, the accurate recognition of road surface types and conditions has emerged as a crucial technology for improving the safety and comfort levels in autonomous driving. This paper puts forward a multi-feature fusion approach for road surface identification. Relying on a 24 GHz millimeter-wave radar, statistical features are combined with wavelet transform techniques. This combination enables the efficient classification of diverse road surface types and conditions. Firstly, the discriminability of radar echo signals corresponding to different road surface types is verified via statistical analysis. During this process, six-dimensional statistical features that display remarkable differences are extracted. Subsequently, a novel radar data reconstruction approach is presented. This method involves fitting discrete echo signals into coordinate curves. Then, discrete wavelet transform is utilized to extract both low-frequency and high-frequency features, thereby strengthening the spatio-temporal correlation of the signals. The low-frequency information serves to capture general characteristics, whereas the high-frequency information reflects detailed features. The statistical features and wavelet transform features are fused at the feature level, culminating in the formation of a 56-dimensional feature vector. Four machine learning models, namely the Wide Neural Network (WNN), K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and Kernel methods, are employed as classifiers for both training and testing purposes. Experiments were executed with 8865 samples obtained from a real-vehicle platform. These samples comprehensively represented 12 typical road surface types and conditions. The experimental outcomes clearly indicate that the proposed method is capable of attaining a road surface type identification accuracy as high as 94.2%. As a result, it furnishes an efficient and cost-efficient road perception solution for intelligent driving systems. This research validates the potential application of millimeter-wave radar in intricate road environments and offers both theoretical underpinning and practical support for the advancement of autonomous driving technology.

To address the challenges of insufficient multi-scale feature representation and limited local geometric modeling in sparse point cloud 3D object detection, this paper proposes an efficient sparse voxel detection network named SAVF-Net (Sparse Attention and Voxel Fusion Network). Built upon a fully sparse detection framework, SAVF-Net introduces two core components: a Neighborhood Sparse Attention (NSA) module and a Parallel Dual-Tower Multi-Scale Voxel Fusion (MSVF-SSM) module. The NSA module leverages a local self-attention mechanism based on sparse neighborhood indices to effectively capture geometric correlations, thereby enhancing the representation of small or boundary objects. Simultaneously, the MSVF-SSM module utilizes parallel sparse convolutions with varying dilation rates, integrated with a channel-spatial dual-attention mechanism, to adaptively fuse global semantic and local geometric features. Experimental results on the KITTI dataset demonstrate that SAVF-Net significantly outperforms mainstream methods, achieving improvements of approximately 2.20% in 3D mAP and 2.14% in BEV mAP. Notably, the model yields a 2%–4% accuracy gain in small-object categories such as pedestrians and cyclists. Furthermore, real-world field tests validate the model’s generalization and stability in complex dynamic scenes. SAVF-Net achieves superior accuracy and robustness while maintaining computational efficiency, providing strong support for autonomous driving perception systems.

The inherent sparsity of LiDAR point cloud data presents a fundamental challenge for 3D object detection. During the feature encoding stage, especially in voxelization, existing methods find it difficult to effectively retain the critical geometric structural information contained in these sparse point clouds, resulting in decreased detection performance. To address this problem, this paper proposes an enhanced 3D object detection framework. It first designs a Structured Voxel Feature Encoder that significantly enhances the initial feature representation through intra-voxel feature refinement and multi-scale neighborhood context aggregation. Second, it constructs a Hybrid-Domain Attention-Guided Sparse Backbone, which introduces a decoupled hybrid attention mechanism and a hierarchical integration strategy to realize dynamic weighting and focusing on key semantic and geometric features. Finally, a Scale-Aggregation Head is proposed to improve the model’s perception and localization capabilities for different-sized objects via multi-level feature pyramid fusion and cross-layer information interaction. Experimental results on the KITTI dataset show that the proposed algorithm increases the mean Average Precision (mAP) by 3.34% compared to the baseline model. Moreover, experiments on a vehicle platform with a lower-resolution LiDAR verify the effectiveness of the proposed method in improving 3D detection accuracy and its generalization ability.