This paper presents a quantum-classical reinforcement learning framework designed to improve exploration, stability, and coordination in multi-UAV systems operating under partial observability and non-stationary dynamics. The method integrates centralized training with decentralized execution and introduces a quantum-inspired optimization layer implemented via simulated variational circuits. This hybrid structure reshapes the exploration landscape through correlated sampling and energybased objective refinement, allowing UAV agents to avoid premature convergence and maintain robust performance under noise and perturbations. A formal mathematical model of the actor, critic, and quantum-inspired parameterization is provided, along with a complexity analysis covering runtime and sample efficiency. Extensive experiments are conducted in cooperative UAV surveillance, multi-aircraft task allocation, and adversarial pursuit–evasion scenarios. Additional physical-world validation is performed on a 3-UAV testbed. Results demonstrate improved convergence stability, reduced sensitivity to hyperparameters, and consistent gains over classical baselines without overstating claims of quantum advantage. All code and experimental scripts are publicly available for full reproducibility.
Taghavi, M. & Vahidi, J. (2026). A Quantum-Classical Framework for UAV-based Multi-Agent Reinforcement Learning. (e742143). Mathematics and Computational Sciences, (), e742143 https://doi.org/10.30511/mcs.2026.2078733.1639
MLA
Taghavi, M., & Vahidi, J. "A Quantum-Classical Framework for UAV-based Multi-Agent Reinforcement Learning" .e742143 , Mathematics and Computational Sciences, , 2026, e742143. doi: 10.30511/mcs.2026.2078733.1639
HARVARD
Taghavi M., Vahidi J. (2026). 'A Quantum-Classical Framework for UAV-based Multi-Agent Reinforcement Learning', Mathematics and Computational Sciences, (), e742143. doi: 10.30511/mcs.2026.2078733.1639
CHICAGO
M. Taghavi & J. Vahidi, "A Quantum-Classical Framework for UAV-based Multi-Agent Reinforcement Learning," Mathematics and Computational Sciences, (2026): e742143, doi: 10.30511/mcs.2026.2078733.1639
VANCOUVER
Taghavi M., Vahidi J. A Quantum-Classical Framework for UAV-based Multi-Agent Reinforcement Learning. MCS. 2026;():e742143. doi: 10.30511/mcs.2026.2078733.1639
