An Approach Based on Quantum Reinforcement Learning for Navigation Problems
Níyazí Furkan Bar, Hasan Yetış, Mehmet Karaköse · 2022 International Conference on Data Analytics for Business and Industry (ICDABI) · 2022
The power of classical computers is still insufficient for deep reinforcement learning (DRL) problems which has large state space. Thanks to entanglement and superposition, quantum computers have a high computational power. The concept of using this high computing power to solve problems that would be difficult for classical computers is fairly common. In this study, a hybrid approach is proposed to take advantage of the benefits of quantum computers. Deep Q-Network (DQN) algorithm for DRL, optimization operations, and storage operations are performed on the classical computer side of this hybrid approach. In the quantum side, a variational quantum circuit (VQC) is proposed. The proposed method is applied to a navigation problem. The proposed approach is evaluated in terms of target success rate, collision (going out) rate. The proposed approach is compared to DRL solutions in the literature for the navigation problem with classical computers. According to the number of parameters used, the proposed approach appears to be successful. As a result, the performance of the proposed approach has been validated.