Enhancing the Dynamic Window Approach: Safer Navigation With Improved Local Goal Selection and Simplified Evaluation Functions
Ziang Lin, Jun Hua Gu, Shengfei Li, Xiangyang Su, Naisi Zhang, Yang Wang · 2025
The Dynamic Window Approach (DWA) is a widely used algorithm for autonomous robot navigation. However, its conventional method for local goal selection often leads to overly aggressive behavior, compromising safety in environments with obstacles. Additionally, the standard evaluation function is often overly complex, containing redundant components that complicate parameter tuning. This paper introduces two new evaluation criteria-goaldist and fit-and a refined local goal selection method. These contributions simplify the evaluation function by addressing redundancy, reduce the complexity of parameter tuning, and promote safer and smoother navigation. Experimental results show that the proposed approach increases the robot's average and minimum distances to obstacles by 0.35 m and 0.6 m, respectively, while significantly improving navigation success rates and ensuring safer, more reliable operation.