Summary
Quantum computers can help solve complex routing problems like delivering goods efficiently, but real devices have noise that disrupts their calculations. The authors developed a smart controller called APGEM that adjusts different error-correcting methods during the learning process to keep the quantum system accurate. Tested on a realistic city delivery scenario, their approach helped the quantum learning method beat simple heuristics and get close to advanced algorithms despite noise. This shows that adapting how errors are fixed based on what the quantum system is learning can make quantum reinforcement learning more practical.
What this means in practice
- •For quantum software developers: Improve the stability and accuracy of quantum reinforcement learning algorithms for complex logistics problems on real noisy quantum devices using adaptive error mitigation.
- •For urban logistics planners: Use quantum-enhanced learning methods with adaptive error mitigation to generate more efficient vehicle routing plans over real geographic data despite hardware noise.
Abstract
Quantum Reinforcement Learning (QRL) represents policies as variational quantum circuits (VQCs), making it attractive for combinatorial optimization such as the Capacitated Vehicle Routing Problem (CVRP). On noisy intermediate-scale quantum (NISQ) hardware, however, decoherence degrades fidelity and destabilizes learning, and conventional error mitigation is applied statically without regard to the learning context. We introduce Adaptive Policy-Guided Error Mitigation (APGEM), a controller that selects among Zero-Noise Extrapolation (ZNE), Probabilistic Error Cancellation (PEC), Clifford Data Regression (CDR), and Readout Error Mitigation (REM) online, driven by a fidelity, entropy, and cost aware utility function and an epsilon-greedy rule over temporal-difference Q-scores. We evaluate on a realistic urban-logistics testbed, a Delhi-based CVRP over real landmarks with geodesic inter-node costs, exercised across five noise families and four severity levels. On this instance, the QRL agent outperforms constructive heuristics and approaches metaheuristics, while mitigation restores approximation ratios from 0.84-0.87 to 0.92-0.94 under high noise. The controller shifts from a CDR-dominated regime under short training horizons to a balanced deployment across all four techniques under longer horizons, indicating genuine regime-dependent selection. These preliminary results position adaptive, learning-aware mitigation as a practical route to noise-resilient QRL.