Towards autonomous tuning of noisy quantum dots

Gate-defined quantum dots (QDs), in which electrons are trapped in quantum wells defined by gate voltages, are a quantum computing platform that may scale effectively due to the maturity of semiconductor processing. However, initialization of these devices is not trivial and currently performed mostly manually or in a semi-scripted fashion guided by heuristics. There has been some progress towards autonomous tuning of these devices using machine learning (ML) methods, but the current best strategies are not robust to ever-present noise in the system or require human-labelled noisy data. We are working to overcome issues of imperfect devices while eliminating the labor of labelling data by incorporating noise into our QD simulator. With this approach, we broaden the applicability of autonomous tuning methods to less ideal devices while using a scalable simulation-based ML framework.