Enhancing DWA via Focus-band Sampling Refinement and Adaptive Parameter Adjustment

This study addresses two limitations of the DWA (Dynamic Window Approach) for local path planning, namely computational inefficiency caused by static velocity-space sampling and local-minima convergence under fixed parameters. To improve efficiency using only simple numerical operations, we propose a FB (Focus-Band) adaptive sampling method, termed FB-DWA, which detects collision clusters in the velocity space and selectively refines candidates near their boundary regions. To further mitigate residual local-minima behavior, we introduce an Adaptive-Parameter method that measures driving difficulty using a collision ratio and switches key parameters across four driving modes. AFB-DWA improves local planning efficiency and robustness by integrating Focus-Band-based candidate refinement with situation-adaptive parameter switching. ROS1 Gazebo simulations show reduced travel time and smoother trajectories with a high success rate, supporting real-time navigation without complex optimization.