Digital-Twin Benchmarks

The simulations include realistic broadening, kinematics, motion costs, and noise, but they are not live instrument deployments. The benchmarks use simplified square-lattice models with at most three exchange parameters to isolate the control architecture. The simulations use realistic thermal-neutron triple-axis assumptions, including fixed final energy, collimations, PG optics, kinematic constraints, and Cooper-Nathans-derived broadening. Synthetic data combine Poisson counting noise with a systematic intensity-dependent floor. Resolution treatment broadens models only in energy rather than using full four-dimensional convolution.