Quantum Reinforcement Learning for Coordinated Satellite Systems

gyu seon kim, Samuel Yen-Chi Chen, Soohyun Park, Joongheon Kim

Keywords: Quantum Reinforcement Learning (QRL), Quantum Neural Network (QNN), Mobility, Satellite Systems

Abstract:

Reinforcement learning (RL) based on classical neural networks (NN) has demonstrated remarkable advancements across diverse domains. Despite this progress, classical RL encounters training difficulties in systems characterized by high-dimensional action spaces, such as coordinated mobility and satellite systems. In these complex settings, the rapid growth in computational resources required due to increased model parameters substantially limits scalability and convergence speed. Quantum reinforcement learning (QRL), which utilizes quantum neural networks (QNN), offers a promising solution by leveraging quantum mechanical properties, such as superposition and entanglement. QNN particularly enables compact representation of multiple states simultaneously using fewer quantum bits (qubits), drastically reducing computational demands. Owing to its distinct features of rapid convergence and enhanced scalability, QRL emerges as a suitable alternative to classical RL approaches for coordinated mobility and satellite applications. Furthermore, the proposed QRL framework effectively alleviates the curse of dimensionality through efficient utilization of qubits.

(We want to have the paper as extended abstract, if accepted)

Submission Number: 16