Safe Control for Learning-Enabled Autonomy with Conformal Prediction

Learning-enabled autonomous systems promise to enable many future technologies such as autonomous driving, intelligent transportation, and robotics. Accelerated by advances in machine learning and AI, there has been tremendous success in the design of learning-enabled autonomous systems. However, these exciting developments are accompanied by new fundamental challenges that arise regarding the safety and reliability of these increasingly complex control systems in which sophisticated algorithms interact with unknown dynamic environments. In this talk, I will provide new insights and discuss exciting opportunities to address these challenges.

I will start by motivating the need for an algorithmic and formal view on systems theory where control algorithms are learning-enabled, formally verified, real-time capable, and reactive to the physical world. Imperfect learning and unknowns in the environment require control techniques to rigorously account for uncertainties. I advocate for the use of conformal prediction (CP) — a statistical tool for uncertainty quantification — due to its simplicity, generality, and efficiency as opposed to existing optimization techniques that are either conservative or not scalable. I first provide an introduction to CP for the non-expert. My goal is then to show how we can use CP to solve the problem of designing probabilistically safe control algorithms in dynamic environments. Specifically, we will design a model predictive controller in conjunction with (i) learning-enabled trajectory predictors to obtain predictions of the environment, and (ii) conformal prediction regions quantifying uncertainty of these predictions. We will also discuss how to deal with high-level temporal logic specifications and distribution shifts that arise when the deployed learning-enabled system changes.