Answers are grounded in this paper and cite their sources.
This letter proposes Spectral-IRA, a data-driven Stochastic Model Predictive Control (SMPC) framework to manage nonlinear Joint Chance Constraints (JCCs) under stochastic uncertainties. To balance computational tractability and risk assessment, we construct a calibrated linear surrogate of the stagewise violation risk that intuitively integrates the physical safety boundary in the state space, a deterministic linear score in the feature space, and the prescribed risk budget in the probability space. By leveraging Kernel Mean Embeddings (KME) to propagate expectations of nonlinear random features that remain sensitive to non-Gaussian distribution shapes, the framework translates intractable probabilistic constraints into structurally convex linear inequalities. To establish empirical safety on the calibration distribution, epistemic approximation errors are accounted for using a fixed offline-calibrated spatial margin. Furthermore, Spectral-IRA adaptively redistributes risk budgets to mitigate conservatism while maintaining nominal recursive feasibility. Simulations demonstrate that our approach satisfies the prescribed risk budgets, offering a practical trade-off among safety, performance, and computational efficiency.
No takes yet. Share an insight, caveat, or question.
Koo et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: