Distributionally Robust Imitation Learning: Layered Control Architecture for Certifiable Autonomy

2025-12-19

AI Summary

Imitation learning (IL) suffers from compounded distributional shift in uncertain dynamical systems, arising jointly from policy errors and model/environment uncertainties—rendering performance unverifiable. To address this, we propose a Hierarchical Control Architecture (LCA) that integrates Taylor-series-based Imitation Learning (TaSIL) with ℓ₁-norm distributionally robust adaptive control (ℓ₁-DRAc), yielding an end-to-end verifiable autonomous closed-loop system. Our key innovation is the Distributionally Robust Inter-layer Policy (DRIP) framework, which enforces input-output constraints across hierarchical layers to jointly guarantee policy robustness and model-uncertainty robustness. By unifying distributionally robust optimization with deterministic certificate generation, LCA achieves certified reliability across the full perception-decision-control pipeline. This work provides the first scalable, theoretically grounded, certifiable IL solution for safety-critical AI systems, with formal performance guarantees under bounded uncertainties.

Technology Category

• Machine Learning: Imitation Learning & Inverse Reinforcement Learning
• Intelligent Robots: Behavior Learning & Control
• Reasoning under Uncertainty: Sequential Decision Making

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Imitation learning (IL) enables autonomous behavior by learning from expert demonstrations. While more sample-efficient than comparative alternatives like reinforcement learning, IL is sensitive to compounding errors induced by distribution shifts. There are two significant sources of distribution shifts when using IL-based feedback laws on systems: distribution shifts caused by policy error and distribution shifts due to exogenous disturbances and endogenous model errors due to lack of learning. Our previously developed approaches, Taylor Series Imitation Learning (TaSIL) and $mathcal{L}_1$ -Distributionally Robust Adaptive Control (ellonedrac), address the challenge of distribution shifts in complementary ways. While TaSIL offers robustness against policy error-induced distribution shifts, ellonedrac offers robustness against distribution shifts due to aleatoric and epistemic uncertainties. To enable certifiable IL for learned and/or uncertain dynamical systems, we formulate extit{Distributionally Robust Imitation Policy (DRIP)} architecture, a Layered Control Architecture (LCA) that integrates TaSIL and~ellonedrac. By judiciously designing individual layer-centric input and output requirements, we show how we can guarantee certificates for the entire control pipeline. Our solution paves the path for designing fully certifiable autonomy pipelines, by integrating learning-based components, such as perception, with certifiable model-based decision-making through the proposed LCA approach.

Problem

• Addresses distribution shifts from policy errors and uncertainties
• Integrates robust imitation learning with adaptive control layers
• Enables certifiable autonomy for learned or uncertain dynamical systems

Innovation

• Integrates TaSIL and L1-DRAC for layered control
• Addresses distribution shifts from policy error and uncertainties
• Guarantees certificates for entire control pipeline via LCA