Methods and Applications of Autonomous Experimentation

In this chapter, the authors use the concept of Autonomous Experiments (AE) as an alternative and more efficient way of data acquisition in inelastic neutron scattering. With AE they entrust the statistical analysis of the acquired data and the best guess for the future data points to an algorithm, while scientists control the progress of the experiment by defining acquisition and cost functions, potential models, priors, or the type of kernels. They propose three different examples, which apply to different types of spectrometers, either in real experiments or in simulations. While all cases use the cyclic concept of Autonomous Experimentation, algorithms, and steering processes differ from case to case. Classical triple-axis spectroscopy (TAS) measurements follow same steps where, up to now, belief and decision policy rely entirely on scientists. Because the data acquisition with TAS is sequential and on the order of minutes, after each measured point, a decision is made about what the next point to measure should be.