A method is proposed for autonomous modeling and optimization of an electric thruster that applies Bayesian optimization on a Gaussian Process Regression model generated in real time from experimental telemetry. The method can be combined with a prescribed objective function and optimization scheme to optimize the thruster for different mission objectives. A notional asteroid-rendezvous mission powered by a Hall effect thruster is considered as an example where the goal of the optimization is to find the propellant gas mixture (argon:krypton:xenon) that minimizes overall mission cost. The results show that a thruster running on a mixture ratio of 11:87:2 benefits from a 7% reduction in total cost compared to the same thruster running on pure xenon. Analysis of the model reveals how the optimal propellant mixture depends strongly on propellant storage technologies, fluctuations in propellant price, and launch costs. Results from this analysis match trends seen in the commercial market with the move to cheaper propellants.
Thoreau et al. (2026) studied this question.