Fig 1.
DynaFusion-Plan Model Overall Architecture Flowchart: Illustrates the collaborative workflow among the multi-sensor data fusion, path planning, and control decision-making modules, as well as their functional framework for achieving path planning and obstacle avoidance in complex dynamic environments.
Fig 2.
Sensor Data Fusion Module: Includes the overall workflow of multi-sensor data acquisition (LiDAR, depth camera, IMU), preprocessing, CNN feature extraction, LIO-SAM data fusion, and real-time dynamic environment updates.
Fig 3.
Path Planning Module: Includes the full workflow of global path generation, local obstacle avoidance (APF), dynamic path adjustment (DDPG), and path smoothing optimization, as well as the collaborative mechanisms between modules.
Fig 4.
Decision-Making and Control Module Architecture: This includes the collaborative workflow of path tracking and real-time optimization based on MPC, along with dynamic obstacle trajectory prediction and path adjustment based on LSTM.
Table 1.
Key Information of TartanAir, NuScenes, and AirSim Datasets.
Table 2.
Summary of Experimental Setup, which presents the hardware configuration, model training parameters, and key settings for the path planning algorithm.
Table 3.
Performance Comparison of the DynaFusion-Plan Model with Other Baseline Models on the TanAir Dataset, NuScenes Dataset and AirSim Dataset.
Fig 5.
Visualization of Comparative Performance Results Across Datasets for Different Path Planning Models.
Table 4.
Ablation Study Results for DynaFusion-Plan Across TartanAir, NuScenes, and AirSim Datasets.
Fig 6.
Visualization of Ablation Study Results for Different Model Configurations on TartanAir, NuScenes, and AirSim Datasets.
Fig 7.
Obstacle Detection and Labeling: Obstacle detection results and initial path labeling in three dynamic environments. It shows how the model accurately detects and labels obstacles in different scenarios, ensuring the effectiveness of subsequent path adjustments.
Fig 8.
Dynamic Path Planning Adjustment Process: The complete process from initial planning to obstacle avoidance adjustment. It demonstrates how the model dynamically adjusts the path in response to changing obstacles, ensuring safety and reachability.
Fig 9.
Final Path Smoothing Optimization: Step-by-step results of smoothing optimization applied to the adjusted path. It shows how the path becomes more continuous and efficient through smoothing optimization, enhancing its feasibility and control performance.