End-to-end QP-based policies: A unified perspective on robust control and robot learning

Fausto Vega; Priyanka Supraja Balaji; Chase Dunaway; Joe Koszut; Jon Arrizabalaga; Zachary Manchester

cs.ROSep 25, 2026

End-to-end QP-based policies: A unified perspective on robust control and robot learning

Authors: Fausto Vega, Priyanka Supraja Balaji, Chase Dunaway, Joe Koszut, Jon Arrizabalaga, Zachary Manchester

Organizations: Department of Aeronautics and Astronautics, Massachusetts Institute of Technology, Cambridge, MA, USA · Department of Mechanical Engineering, California Institute of Technology, Pasadena, CA, USA

Abstract

We present an end-to-end QP-based policy framework that enables systematic policy construction with minimal domain-specific design, while preserving the transparency and interpretability of model-based control. The proposed policy representation supports both domain-randomized model-based auto-tuning, where policy parameters are optimized over distributions of disturbances and model variations, and black-box policy construction, where the problem is formulated in terms of a (possibly) unknown model without requiring explicit notions of states, inputs, or the underlying system dynamics. We establish connections to existing policy representations and control paradigms, including robust control, multilayer perceptrons, and robot learning, and interpret our end-to-end QP policies as a common abstraction of these approaches. We validate the resulting framework in both simulation and hardware, demonstrating a broad range of applications spanning robustness, automatic policy tuning, and control under unknown system dynamics.

Figures & tables

Panel of Figure 1 — Fig. 1 : Robust-control connections of the proposed QP policy on a double integrator regulation task. (a) Disturbance uncertainty: optimizing the outer objective approximates $H_{2}$ - and $H_{\infty}$ -type feedback gains. (b) Model uncertainty: entropic risk interpolates between average- and worst-case performance. (c) Combined model and disturbance uncertainty: the learned QP policy is compared with nominal $H_{\infty}$ and worst case $H_{\infty}$ under plant uncertainty.

Fig. 2 : Percentage of landings within a specified landing radius for 1000 vehicles, with and without domain randomization (DR). The inset highlights the upper landing-error tail.

Fig. 3 : Closed-loop phase portraits of the learned projection-QP policy and the single-hidden-layer ReLU policy from the same five initial conditions near the hanging equilibrium. Both policies generate swing-up trajectories and converge to the upright equilibrium.

Fig. 4 : Our end-to-end QP policy applied to a Crazyflie drone under a $5\ \mathrm{m/s}$ crosswind generated by an industrial fan with a flow-conditioning rack. Estimator, offset-compensation, and cost parameters are learned end to end, removing manual tuning and outperforming the hand-tuned baselines.

Fig. 5 : Closed-loop position error under crosswind for the learned end-to-end QP policy and two manually tuned MPC baselines based on Bryson’s rule. Thick curves show the mean over five trials, thin transparent curves show individual trials, and light shading marks the trial range.

Robot Policy Learning

Robot policy learning develops controllers that map sensory observations, task goals, and interaction history to actions for embodied robots.

See all 372 papers

Learning-Based Robust Control: Unifying Exploration and Distributional Robustness for Reliable Robotics via Free Energy
Mar 6, 2026Hozefa Jesawada, Giovanni Russo, Abdalla Swikir +1Robust Control Distributionally Robust RL
Mechanical Engineering Program, New York University Abu Dhabi, Abu Dhabi, UAE. · Department of Information and Electrical Engineering & Applied Mathematics, University of Salerno, Fisciano, Italy. · Mohammed bin Zayed University for Artificial Intelligence, Abu Dhabi, UAE.
Finite-Time Convergence of Single-Trajectory Chi-Square Robust Q-Learning With Linear Function Approximation
Oct 2, 2025Saptarshi Mandal, Yashaswini Murthy, R. SrikantQ-Learning Distributionally Robust RL
ECE, UIUC · ORIE, UT Austin · ECE, CSL, and the director of NCSA at UIUC
Continual Robot Policy Learning via Variational Neural Dynamics
Jun 25, 2026Jiaxu Xing, Zhiyuan Zhu, Yunfan Ren +4Robot Policy Learning Robot Policy Adaptation
Robotics and Perception Group, University of Zurich

Mar 6, 2026J/K move · Enter open · S save

Learning-Based Robust Control: Unifying Exploration and Distributional Robustness for Reliable Robotics via Free Energy

Hozefa Jesawada, Giovanni Russo, Abdalla Swikir, Fares Abu-Dakka

Mechanical Engineering Program, New York University Abu Dhabi, Abu Dhabi, UAE. · Department of Information and Electrical Engineering & Applied Mathematics, University of Salerno, Fisciano, Italy. · Mohammed bin Zayed University for Artificial Intelligence, Abu Dhabi, UAE.

Robust Control Distributionally Robust RL Free Energy Principle Model-Based RL Diffusion Models Robotic Manipulation

Open paper AskPDF arXiv

A key challenge towards reliable robotic control is devising computational models that can both learn policies and guarantee robustness when deployed in the field. Inspired by the free energy principle in computational neuroscience, to address these challenges, we propose a model for policy computation that jointly learns environment dynamics and rewards, while ensuring robustness to epistemic uncertainties. Expounding a distributionally robust free energy principle, we propose a modification to the maximum diffusion learning framework. After explicitly characterizing robustness of our policies to epistemic uncertainties in both environment and reward, we validate their effectiveness on continuous-control benchmarks, via both simulations and real-world experiments involving manipulation with a Franka Research~3 arm. Across simulation and zero-shot deployment, our approach narrows the sim-to-real gap, and enables repeatable tabletop manipulation without task-specific fine-tuning.

Abstract

Robot Policy Learning

Explore similar work

Learning-Based Robust Control: Unifying Exploration and Distributional Robustness for Reliable Robotics via Free Energy

Finite-Time Convergence of Single-Trajectory Chi-Square Robust Q-Learning With Linear Function Approximation

Continual Robot Policy Learning via Variational Neural Dynamics