A Predictive and Risk-Adaptive Dynamic Window Approach for Robot Navigation in Constrained Environments
Reliable robot navigation in spatially constrained dynamic environments requires coordinated global guidance, local planning, and obstacle prediction. This study proposes a predictive and risk-adaptive Dynamic Window Approach (DWA) to address fixed evaluation weights and limited anticipation of obstacle motion. An asymmetric feasible region is constructed along the global reference path from inflated static obstacles and map boundaries to constrain local targets and candidate trajectories. Robot trajectories and obstacle motions are assessed at synchronized future instants. Predicted safety margins support trajectory screening, while a composite risk measure incorporating approximate time-to-collision adjusts evaluation weights. Passage decisions coordinate waiting, safe traversal, and global replanning under temporary or persistent blockage. Simulations included mechanism validation and comparisons with five DWA-based baselines. The comprehensive experiment comprised 30 randomized trials per method. The proposed method achieved an 80.0% success rate, compared with 56.7% for Time-Aligned Prediction (TAP). Among successful collision-free runs, the proposed method achieved a mean tracking error of 0.269 m, a mean minimum dynamic safety margin of 0.541 m, and a mean arrival time of 75.80 s; TAP’s corresponding mean arrival time was 56.86 s. These results indicate a trade-off between task-completion robustness and traversal efficiency under the tested conditions.
Authors
- Yongtao Qiu (ORCID: https://orcid.org/0000-0002-8993-4277)
- Ruhong Ma
- Yong Zeng (ORCID: https://orcid.org/0000-0003-2401-416X)
- Zheng Zhang
Institutions
- Yancheng Institute of Technology (CN)
Publication Details
- Journal
- Processes
- Published
- 2026-11-18
- DOI
- https://doi.org/10.3390/pr14223706
- Primary Topic
- Robotic Path Planning Algorithms
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- Yancheng Institute of Technology