To address the navigation challenges of unmanned ground vehicles (UGVs) in unknown, unstructured environments, we propose the Dynamic Weighted Perception Avoidance Algorithm (DWPA). DWPA features LiDAR-based real-time perception and a dynamic weighting mechanism that autonomously balances goal-directed navigation with obstacle avoidance. A key innovation is the introduction of a hysteresis factor with nonlinear low-pass filtering properties, which suppresses oscillatory control commands triggered by perceptual jumps and enhances robustness. Theoretical reliability is established through Lyapunov stability analysis and control barrier function-based safety proofs. Co-simulations (MATLAB/V-REP) and physical experiments validate DWPA against the artificial potential field (APF) and model predictive control (MPC) methods. Results show that DWPA generates smoother, safer trajectories than MPC while maintaining a larger safety margin, and reduces path length by over 14.4% compared to APF. Its computational efficiency and rapid response make it particularly effective for complex, dynamically changing terrains.
Huang et al. (Sun,) studied this question.