Robot Dynamic Environment Path Planning Based on Kalman-filter-based Prediction and Complexity-Adaptive Collaboration

To tackle the critical challenges confronted by mobile robots in dynamic environments, such as the demand for short replanning time, high computational complexity, and unreliable planning results arising from environmental uncertainty, the majority of existing methods merely implement local obstacle avoidance in the post-processing phase, which leads to limited improvements in global path optimization. Moreover, the fixed coordination mechanism between global and local planning is difficult to adapt to varying scenario pressures. To this end, this paper proposes a planning method with a predictive dynamic cost field and a complexity-adaptive coordination mechanism (PADP), which employs Kalman-filter-based prediction to construct a direction-aware dynamic cost field and integrates it into global path generation. An online complexity indicator composed of obstacle density and other metrics is designed to dynamically adjust planning weights. Meanwhile, a lightweight residual correction module is introduced to reduce multi-step prediction errors while retaining the core Kalman filtering framework. Experimental results based on MATLAB grid simulations demonstrate that PADP achieves an overall success rate of 80.56% and has the minimum average collision count among all comparison methods in 36 dynamic scenarios. The hybrid PADP-KFRes reduces the prediction RMSE from 1.053 to 0.457.