Effective path planning is crucial for autonomous mobile robots navigating complex environments. The “global–local” coupled path planning algorithm exhibits superior global planning capabilities and local adaptability. However, these algorithms often fail to fully realize their potential due to low efficiency and excessive constraints. To address these issues, this study introduces a simpler and more effective integration strategy. Specifically, this paper proposes using a bi-layer map and a feasible domain strategy to organically combine the D*Lite algorithm with the Dynamic Window Approach (DWA). The bi-layer map effectively reduces the number of nodes in global planning, enhancing the efficiency of the D*Lite algorithm. The feasible domain strategy decreases constraints, allowing the local algorithm DWA to utilize its local planning capabilities fully. Moreover, the cost functions of both the D*Lite algorithm and DWA have been refined, enabling the fused algorithm to cope with more complex environments. This paper conducts simulation experiments across various settings and compares our method with A_DWA, another “global–local” coupled approach, which combines A* and DWA. D_DWA significantly outperforms A_DWA in complex environments, despite a 7.43% increase in path length. It reduces the traversal of risk areas by 71.95%, accumulative risk by 80.34%, global planning time by 26.98%, and time cost by 35.61%. Additionally, D_DWA outperforms the A_Q algorithm, a coupled approach validated in real-world environments, which combines A* and Q-learning, achieving reductions of 1.34% in path length, 67.14% in traversal risk area, 78.70% in cumulative risk, 34.85% in global planning time, and 37.63% in total time cost. The results demonstrate the superiority of our proposed algorithm in complex scenarios.

Lane detection is a critical component of autonomous driving systems, requiring both high accuracy and real-time performance under complex driving scenarios. Unlike current methods that rely on predefined lane counts, instance segmentation methods can handle an arbitrary number of lanes, making them more adaptable in real-world applications. However, this flexibility typically relies on dense pixel-level predictions, which necessitate large-scale networks and result in prohibitively high computational costs, hindering deployment on embedded platforms. To address these challenges, we present Fast-CenLaneNet, a lightweight architecture that improves inference efficiency while maintaining detection accuracy. Specifically, we design a lightweight backbone to reduce model parameters and computational cost, propose a learnable spatial similarity attention module to capture spatial dependencies within lane regions and enhance feature discriminability, and construct multi-branch output heads with Ghost convolutions to refine lane-related features with low computational overhead. Experiments on the TuSimple and CULane benchmarks demonstrate that Fast-CenLaneNet achieves a favorable accuracy–efficiency trade-off. On TuSimple, Fast-CenLaneNet obtains 96.40 ± 0.06% accuracy and 162.7 ± 6.8 FPS with 4.7 M parameters and 9.9 GFLOPs. Compared with CenLaneNet, it reduces the number of parameters by 89.1% and improves forward inference speed by 107.5%, with an accuracy decrease of only 0.08 percentage points.