Safe Optimal Control Using Stochastic Barrier Functions and Deep Forward-Backward SDEs
This work addresses safety constraints in stochastic control for applications like robotics or autonomous systems, presenting a novel integration of methods rather than a fundamental breakthrough.
The paper tackles the problem of ensuring safety in stochastic optimal control by introducing a new formulation that combines Forward-Backward Stochastic Differential Equations, Stochastic Barrier Functions, Differentiable Convex Optimization, and Deep Learning, resulting in a neural network architecture for safe trajectory optimization validated through simulations on three systems.
This paper introduces a new formulation for stochastic optimal control and stochastic dynamic optimization that ensures safety with respect to state and control constraints. The proposed methodology brings together concepts such as Forward-Backward Stochastic Differential Equations, Stochastic Barrier Functions, Differentiable Convex Optimization and Deep Learning. Using the aforementioned concepts, a Neural Network architecture is designed for safe trajectory optimization in which learning can be performed in an end-to-end fashion. Simulations are performed on three systems to show the efficacy of the proposed methodology.