Answers are grounded in this paper and cite their sources.
Addressing the challenges of intelligent obstacle avoidance for electric vertical take-off and landing (eVTOL) aircraft in complex urban air mobility (UAM) scenarios, and motivated by the need to support future certification-oriented explainability analysis, this paper proposes an intelligent obstacle avoidance method based on explainable reinforcement learning. First, an explainable-by-design intelligent obstacle avoidance strategy is developed. This strategy centers on the Soft Actor-Critic (SAC) algorithm, which is integrated with the Artificial Potential Field (APF) method to construct a comprehensive state space and a composite reward function. This design enables effective obstacle avoidance and path planning within complex 3D UAM environments. Second, an explainability framework is established by introducing the XGBoost model and the SHapley Additive exPlanations (SHAP) method. This framework maps the learned obstacle avoidance action strategies into a human-understandable semantic space, facilitating feature-level decision attribution and explainability analysis. Finally, the effectiveness of the proposed obstacle avoidance model and the explainability method is evaluated through extensive simulation experiments conducted in complex 3D UAM scenarios. Both qualitative and quantitative analyses are performed across four critical dimensions: fidelity, completeness, consistency, and feature attribution scores. The results show that the proposed method satisfies the path planning and obstacle avoidance requirements while constructing a “decision-explainability” dual link framework with fidelity R² > 0.99 . It provides methodological support for reinforcement-learning-based eVTOL obstacle avoidance and offers simulation-based explainability evidence that may support future airworthiness-oriented analysis of intelligent methods.
No takes yet. Share an insight, caveat, or question.
Dong et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: