Decisions with probabilistic guarantee of constraint satisfaction
Probablistic constraint satisfaction
Chance-constrained optimization incorporates probabilistic constraints into optimization problems. It ensures that the probability of satisfying a constraint is above a specified threshold, allowing decision-makers to account for uncertainty and manage risk in their optimization models.
$$Pr\{g(x,\xi) \le 0 \} \ge 1-\epsilon $$
Our research contribution focuses on addressing the challenges in solving chance constrained optimization problems. The first contribution is the development of a novel algorithm that finds the optimal robust optimization approximation, considering the smallest possible uncertainty set size. The second contribution explores a tractable robust optimization framework for joint chance constrained problems. Additionally, the use of neural network-based approaches, including a ReLU artificial neural network and a recurrent neural network, is proposed to efficiently solve nonlinear joint chance constrained optimization problems in specific applications.
Relevant publications:
S. Yang, Z. Li. Recurrent Neural Network-Based Joint Chance Constrained Stochastic Model Predictive Control. The 13th IFAC Symposium on Dynamics and Control of Process Systems, including Biosystems (DYCOPS), Busan, Korea, 2022.
S. Yang, J. Moreira, Z. Li. Joint Chance Constrained Process Optimization through Neural Network Approximation. The 14th International Symposium on Process Systems Engineering, Kyoto, Japan, 2022.
