SARL: Structure-Aligned Reinforcement Learning for Bridging the Perception-Action Gap in Airspace

Binhao Gu, Jinjun Cai, Weihuang Zheng, Jiaxing Li, Youyong Kong, Hui Ding
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:36942-36959, 2026.

Abstract

Multi-Agent Reinforcement Learning (MARL) has been widely applied to automated aircraft conflict resolution due to its strong capability for cooperative control and distributed decision-making. However, existing approaches typically assume a fixed number of aircraft and neglect the unique characteristics of air traffic control instructions. This structural misalignment between model architectures and domain requirements leads to severe deficiencies in perception scalability and action stability across scenarios of varying scales. To address these challenges, we propose Structural-Aligned Reinforcement Learning (SARL), which aims to bridge the gap between perception and action. First, the Physics-Encoded Relational Graph (PERG) effectively resolves the fixed input dimensionality issue by incorporating physical inductive biases into a graph attention mechanism. Second, we design the Sparse Cognitive Mixture-of-Experts (SC-MoE) to enhance decision stability. In addition, we introduce a Kinematic Kafety Shield (KSS) based on aviation rules, which not only improves inference-time safety but also effectively guides the model to generate semantically meaningful actions that comply with aviation standards. Simulation experiment results demonstrate that SARL significantly outperforms existing reinforcement learning baselines across diverse scenarios in terms of both success rate and operational efficiency.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-gu26a, title = {{SARL}: Structure-Aligned Reinforcement Learning for Bridging the Perception-Action Gap in Airspace}, author = {Gu, Binhao and Cai, Jinjun and Zheng, Weihuang and Li, Jiaxing and Kong, Youyong and Ding, Hui}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {36942--36959}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/gu26a/gu26a.pdf}, url = {https://proceedings.mlr.press/v306/gu26a.html}, abstract = {Multi-Agent Reinforcement Learning (MARL) has been widely applied to automated aircraft conflict resolution due to its strong capability for cooperative control and distributed decision-making. However, existing approaches typically assume a fixed number of aircraft and neglect the unique characteristics of air traffic control instructions. This structural misalignment between model architectures and domain requirements leads to severe deficiencies in perception scalability and action stability across scenarios of varying scales. To address these challenges, we propose Structural-Aligned Reinforcement Learning (SARL), which aims to bridge the gap between perception and action. First, the Physics-Encoded Relational Graph (PERG) effectively resolves the fixed input dimensionality issue by incorporating physical inductive biases into a graph attention mechanism. Second, we design the Sparse Cognitive Mixture-of-Experts (SC-MoE) to enhance decision stability. In addition, we introduce a Kinematic Kafety Shield (KSS) based on aviation rules, which not only improves inference-time safety but also effectively guides the model to generate semantically meaningful actions that comply with aviation standards. Simulation experiment results demonstrate that SARL significantly outperforms existing reinforcement learning baselines across diverse scenarios in terms of both success rate and operational efficiency.} }
Endnote
%0 Conference Paper %T SARL: Structure-Aligned Reinforcement Learning for Bridging the Perception-Action Gap in Airspace %A Binhao Gu %A Jinjun Cai %A Weihuang Zheng %A Jiaxing Li %A Youyong Kong %A Hui Ding %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-gu26a %I PMLR %P 36942--36959 %U https://proceedings.mlr.press/v306/gu26a.html %V 306 %X Multi-Agent Reinforcement Learning (MARL) has been widely applied to automated aircraft conflict resolution due to its strong capability for cooperative control and distributed decision-making. However, existing approaches typically assume a fixed number of aircraft and neglect the unique characteristics of air traffic control instructions. This structural misalignment between model architectures and domain requirements leads to severe deficiencies in perception scalability and action stability across scenarios of varying scales. To address these challenges, we propose Structural-Aligned Reinforcement Learning (SARL), which aims to bridge the gap between perception and action. First, the Physics-Encoded Relational Graph (PERG) effectively resolves the fixed input dimensionality issue by incorporating physical inductive biases into a graph attention mechanism. Second, we design the Sparse Cognitive Mixture-of-Experts (SC-MoE) to enhance decision stability. In addition, we introduce a Kinematic Kafety Shield (KSS) based on aviation rules, which not only improves inference-time safety but also effectively guides the model to generate semantically meaningful actions that comply with aviation standards. Simulation experiment results demonstrate that SARL significantly outperforms existing reinforcement learning baselines across diverse scenarios in terms of both success rate and operational efficiency.
APA
Gu, B., Cai, J., Zheng, W., Li, J., Kong, Y. & Ding, H.. (2026). SARL: Structure-Aligned Reinforcement Learning for Bridging the Perception-Action Gap in Airspace. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:36942-36959 Available from https://proceedings.mlr.press/v306/gu26a.html.

Related Material