Neuro-Symbolic Deadlock Resolution in Multi-Robot Systems

Ruiyang Wang, Bowen He, Miroslav Pajic
Proceedings of the 7th Annual Learning for Dynamics \& Control Conference, PMLR 283:1066-1077, 2025.

Abstract

This work addresses the problem of deadlock situations that are common in decentralized multi-robot missions. Existing approaches focus on predicting potential deadlocks and intervening before they actually occur. However, these methods often struggle to detect all possible deadlocks, especially in environments with uncontrollable obstacles, and may inevitably introduce new dead-locks after interventions. Consequently, we propose a neuro-symbolic deadlock resolution (NSDR) method based on Neural Logic Machines (NLMs). NSDR is designed specifically to resolve dead-locks after their occurrence, with the guarantee that no further persistent deadlocks will emerge after the initial resolution in environments with or without obstacles. Our approach leverages the similarity in logic rules when resolving simple deadlocks involving a small number of robots; this facilitates their use when resolving more complex scenarios with larger robot groups. Training NSDR on simpler deadlock cases allows it to generalize and effectively resolve more complex situations by utilizing the logic rules it has learned from simple deadlocks. We thoroughly evaluate the method in case studies with varying numbers of robots involved in deadlocks and show that NSDR outperforms the state of the art methods, which are based on the use of the adaptive repulsive force.

Cite this Paper

BibTeX
@InProceedings{pmlr-v283-wang25f, title = {Neuro-Symbolic Deadlock Resolution in Multi-Robot Systems}, author = {Wang, Ruiyang and He, Bowen and Pajic, Miroslav}, booktitle = {Proceedings of the 7th Annual Learning for Dynamics \& Control Conference}, pages = {1066--1077}, year = {2025}, editor = {Ozay, Necmiye and Balzano, Laura and Panagou, Dimitra and Abate, Alessandro}, volume = {283}, series = {Proceedings of Machine Learning Research}, month = {04--06 Jun}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v283/main/assets/wang25f/wang25f.pdf}, url = {https://proceedings.mlr.press/v283/wang25f.html}, abstract = {This work addresses the problem of deadlock situations that are common in decentralized multi-robot missions. Existing approaches focus on predicting potential deadlocks and intervening before they actually occur. However, these methods often struggle to detect all possible deadlocks, especially in environments with uncontrollable obstacles, and may inevitably introduce new dead-locks after interventions. Consequently, we propose a neuro-symbolic deadlock resolution (NSDR) method based on Neural Logic Machines (NLMs). NSDR is designed specifically to resolve dead-locks after their occurrence, with the guarantee that no further persistent deadlocks will emerge after the initial resolution in environments with or without obstacles. Our approach leverages the similarity in logic rules when resolving simple deadlocks involving a small number of robots; this facilitates their use when resolving more complex scenarios with larger robot groups. Training NSDR on simpler deadlock cases allows it to generalize and effectively resolve more complex situations by utilizing the logic rules it has learned from simple deadlocks. We thoroughly evaluate the method in case studies with varying numbers of robots involved in deadlocks and show that NSDR outperforms the state of the art methods, which are based on the use of the adaptive repulsive force.} }
Endnote
%0 Conference Paper %T Neuro-Symbolic Deadlock Resolution in Multi-Robot Systems %A Ruiyang Wang %A Bowen He %A Miroslav Pajic %B Proceedings of the 7th Annual Learning for Dynamics \& Control Conference %C Proceedings of Machine Learning Research %D 2025 %E Necmiye Ozay %E Laura Balzano %E Dimitra Panagou %E Alessandro Abate %F pmlr-v283-wang25f %I PMLR %P 1066--1077 %U https://proceedings.mlr.press/v283/wang25f.html %V 283 %X This work addresses the problem of deadlock situations that are common in decentralized multi-robot missions. Existing approaches focus on predicting potential deadlocks and intervening before they actually occur. However, these methods often struggle to detect all possible deadlocks, especially in environments with uncontrollable obstacles, and may inevitably introduce new dead-locks after interventions. Consequently, we propose a neuro-symbolic deadlock resolution (NSDR) method based on Neural Logic Machines (NLMs). NSDR is designed specifically to resolve dead-locks after their occurrence, with the guarantee that no further persistent deadlocks will emerge after the initial resolution in environments with or without obstacles. Our approach leverages the similarity in logic rules when resolving simple deadlocks involving a small number of robots; this facilitates their use when resolving more complex scenarios with larger robot groups. Training NSDR on simpler deadlock cases allows it to generalize and effectively resolve more complex situations by utilizing the logic rules it has learned from simple deadlocks. We thoroughly evaluate the method in case studies with varying numbers of robots involved in deadlocks and show that NSDR outperforms the state of the art methods, which are based on the use of the adaptive repulsive force.
APA
Wang, R., He, B. & Pajic, M.. (2025). Neuro-Symbolic Deadlock Resolution in Multi-Robot Systems. Proceedings of the 7th Annual Learning for Dynamics \& Control Conference, in Proceedings of Machine Learning Research 283:1066-1077 Available from https://proceedings.mlr.press/v283/wang25f.html.

Related Material