Journal papers
Lukas Molnar, Jin Cheng, Gabriele Fadini, Dongho Kang,
Fatemeh Zargarbashi, Stelian Coros
IEEE Robotics and Automation Letters (2025 to appear)
SI: Advancements in MPC and Learning Algorithms for Legged Robots
Abstract
Loco-manipulation demands coordinated whole-body motion to manipulate objects effectively while maintaining locomotion stability, presenting significant challenges for both planning and control. In this work, we propose a whole-body model predictive control (MPC) framework that directly optimizes joint torques through full-order inverse dynamics, enabling unified motion and force planning and execution within a single predictive layer. This approach allows emergent, physically consistent whole-body behaviors that account for the system’s dynamics and physical constraints. We implement our MPC formulation using open software frameworks (Pinocchio and CasADi), along with the state-of-the-art interior-point solver Fatrop. In real-world experiments on a Unitree B2 quadruped equipped with a Unitree Z1 manipulator arm, our MPC formulation achieves real-time performance at 80 Hz. We demonstrate loco-manipulation tasks that demand fine control over the end-effector’s position and force to perform real-world interactions like pulling heavy loads, pushing boxes, and wiping whiteboards.
Jin Cheng, Dongho Kang, Gabriele Fadini, Guanya Shi, Stelian Coros
IEEE Robotics and Automation Letters (2025/8/1)
Abstract
Loco-manipulation, physical interaction of various objects that is concurrently coordinated with locomotion, remains a major challenge for legged robots due to the need for both precise end-effector control and robustness to unmodeled dynamics. While model-based controllers provide precise planning via online optimization, they are limited by model inaccuracies. In contrast, learning-based methods offer robustness, but they struggle with precise modulation of interaction forces. We introduce RAMBO, a hybrid framework that integrates model-based whole-body control within a feedback policy trained with reinforcement learning. The model-based module generates feedforward torques by solving a quadratic program, while the policy provides feedback corrective terms to enhance robustness. We validate our framework on a quadruped robot across a diverse set of real-world loco-manipulation tasks, such as pushing a shopping cart, balancing a plate, and holding soft objects, in both quadrupedal and bipedal walking. Our experiments demonstrate that RAMBO enables precise manipulation capabilities while achieving robust and dynamic locomotion.
E. Mingo Hoffman; D. Costanzi; G. Fadini; N. Miguel; A. Del Prete; L. Marchionni
IEEE Robotics and Automation Letters ( Volume: 10, Issue: 3, March 2025)
Abstract
Designing robotic manipulators for generic tasks while meeting specific requirements is a complex, iterative process involving mechanical design, simulation, control, and testing. New computational design tools are needed to simplify and speed up such processes. This work presents an original formulation of the computational design problem, tailored to help design generic manipulators with strong reachability requirements. The primary challenges addressed in this work are twofold. First, the necessity to consider the design of both continuous quantities and discrete components. Second, the ability to guide the design using high-level requirements, like the robot's workspace, without needing a specific manipulation task, unlike other co-design frameworks. These two challenges are addressed by employing a novel kineto-static formulation, resulting in a Mixed Integer Nonlinear Programming problem, which is solved using bi-level optimization. A compelling use case from a real industrial application is presented to highlight the practical effectiveness of the proposed method.
Gabriele Fadini, Shivesh Kumar, Rohit Kumar, Thomas Flayols, Andrea del Prete, Justin Carpentier, Philippe Souères
Cambridge Robotica, 9 May 2024
Abstract
This paper presents a bi-level optimization framework to concurrently optimize a quadruped hardware and control policies for achieving dynamic cyclic behaviors. The longterm vision to drive the design of dynamic and efficient robots by means of computational techniques is applied to improve the development of a new quadruped prototype. The scale of the robot and its actuators are optimized for energy efficiency considering a complete model of the motor, that includes friction, torque, and bandwidth limitations. This model is used to optimize the power consumption during bounding and backflip tasks and is validated by tracking the output trajectories on the first prototype iteration. The co-design results show an improvement of up to 87% for a single task optimization. It appears that, for jumping forward, robots with longer thighs perform better, while for backflips, longer shanks are better suited. To understand the trade-off between these different choices, a Pareto set is constructed to guide the design of the next prototype.
Gabriele Fadini, Thomas Flayols, Andrea del Prete, Philippe Souères
IEEE Robotics and Automation Letters ( Volume: 7, Issue: 4, October 2022)
Abstract
This paper outlines a bi-level optimization method to concurrently optimize robot hardware parameters and control trajectories that ensure robust performance. The outer loop consists in a genetic algorithm that optimizes the hardware according to its average performance when tracking a locally optimal trajectory in perturbed simulations. The tracking controller exploits the locally optimal feedback gains computed in the inner loop with a Differential Dynamic Programming algorithm, which also finds the optimal reference trajectories. Our simulations feature a complete actuation model, including friction compensation and bandwidth limits. Our method can potentially account for arbitrary perturbations, and it discards hardware designs that cannot robustly track the reference trajectories. Our results show improved performance of the designed platform in realistic application scenarios, autonomously leading to the selection of lightweight and more transparent hardware.
Conference papers
Gabriele Fadini, Deepak Ingole, Tong Duy Son, Alisa Rupenyan
2026 IEEE European Control Conference (ECC)
Abstract
This paper presents an auto-tuning framework for torque-based Nonlinear Model Predictive Control (nMPC), where the MPC serves as a real-time controller for optimal joint torque commands. The MPC parameters, including cost function weights and low-level controller gains, are optimized using high-dimensional Bayesian Optimization (BO) techniques, specifically Sparse Axis-Aligned Subspace (SAASBO) with a digital twin (DT) to achieve precise end-effector trajectory real-time tracking on an UR10e robot arm. The simulation model allows efficient exploration of the high-dimensional parameter space, and it ensures safe transfer to hardware. Our simulation results demonstrate significant improvements in tracking performance (+41.9%) and reduction in solve times (-2.5%) compared to manually-tuned parameters. Moreover, experimental validation on the real robot follows the trend (with a +25.8% improvement), emphasizing the importance of digital twin-enabled automated parameter optimization for robotic operations.
Gabriele Fadini, Stelian Coros
2025 IEEE International Conference on Robotics and Automation (ICRA 2025),
May 2025, Atlanta, US
Abstract
We present a novel approach to quantifying and optimizing stability in robotic systems based on the Lyapunov exponents addressing an open challenge in the field of robot analysis, design, and optimization. Our method leverages differentiable simulation over extended time horizons. The proposed metric offers several properties, including a natural extension to limit cycles commonly encountered in robotics tasks and locomotion. We showcase, with an ad-hoc JAX gradient-based optimization framework, remarkable power, and flexibility in tackling the robustness challenge. The effectiveness of our approach is tested through diverse scenarios of varying complexity, encompassing high-degree-of-freedom systems and contact-rich environments. The positive outcomes across these cases highlight the potential of our method in enhancing system robustness.
G. Fadini, T. Flayols, A. Del Prete, N. Mansard, P. Souères
IEEE International Conference on Robotics and Automation (ICRA 2021),
May 2021, Xian, China
Abstract
This paper presents a computational framework for the design of high-performance legged robotic systems. The framework relies on the concurrent optimization of hardware parameters and control trajectories to find the best robot design for a given task. In particular, we focus on energy efficiency, presenting novel electro-mechanical models to account for the losses of the actuators due to friction and Joule effects. Thanks to a bi-level optimization scheme, featuring a genetic algorithm in the outer loop, our framework can also optimize for the duration of the motion, the actuators, and the size of the robot. We present a novel approach to scale both the actuators and the robot structure in a way that ensures structural integrity by maintaining constant the normalized deflection of the links. We validated our approach by designing a two-joint monoped robot to execute a jumping task. Our results show that our framework can lead to remarkable energy savings (up to 60%) thanks to the concurrent optimization of robot size, motion duration, and actuators.
Preprints
Dyuman Aditya, Jin Cheng, Clemens Schwarke, Quan Nguyen, Gaurav Sukhatme, Stelian Coros, Gabriele Fadini
Submitted to ICRA 2027
Abstract
Differentiable simulation provides analytic gradients of robot dynamics, enabling fast and sample-efficient first-order policy optimization. However, obtaining smooth and informative gradients through rigid-body contact typically requires softened contact models, often at the expense of physical fidelity and thereby limiting learned policies largely to simulation. This trade-off becomes particularly consequential for dynamic humanoid motions, where accurate contact dynamics are critical for transferring policies to the real world. Increasing contact stiffness in rigid-body simulation improves the fidelity of interactions, but also makes the dynamics increasingly sensitive to small state perturbations, producing high-variance gradients that can destabilize first-order policy learning. To address this, we propose Bundled Contact Gradients (BCG), a contactlocal randomized smoothing framework for differentiable policy learning. When stiff contact is detected, our method evaluates a local bundle of randomized perturbation rollouts around the stiff contact configuration and aggregates their gradient signal thereby reducing gradient variance. We demonstrate the effectiveness of our method by successfully training and transferring dynamic motions zero-shot onto a real-world Unitree G1 humanoid platform.
Severin Bochem, Eduardo Gonzalez-Sanchez, Yves Bicker, Gabriele Fadini
Abstract
Reinforcement learning often requires extensive training data. Simulation-to-real transfer offers a promising approach to address this challenge in robotics. While differentiable simulators offer improved sample efficiency through exact gradients, they can be unstable in contact-rich environments and may lead to poor generalization. This paper introduces a novel approach integrating sharpness-aware optimization into gradient-based reinforcement learning algorithms. Our simulation results demonstrate that our method, tested on contact-rich environments, significantly enhances policy robustness to environmental variations and action perturbations while maintaining the sample efficiency of first-order methods. Specifically, our approach improves action noise tolerance compared to standard first-order methods and achieves generalization comparable to zeroth-order methods. This improvement stems from finding flatter minima in the loss landscape, associated with better generalization. Our work offers a promising solution to balance efficient learning and robust sim-to-real transfer in robotics, potentially bridging the gap between simulation and real-world performance.
Workshops
Gabriele Fadini, Stelian Coros
Workshop "Morphology-Aware Policy and Design Learning Workshop (MAPoDeL)" at CoRL 2024, Munich, Germany
Abstract
Our ongoing research aims to investigate the potential of integrating robot design optimization with reinforcement learning (RL). In co-design literature, exploiting the ties between design and control is the key to unlocking otherwise unreachable performance. However, the problem of obtaining optimal policies that will adapt to a range of different robots is still open. We reason about the challenges that this problem setting brings. Moreover, we hint at a few possible future research directions that may help advance robot morphology and design-aware control policies while ensuring their optimality with respect to the task.
Gabriele Fadini, Stelian Coros
Workshop "Advancements in Trajectory Optimization and Model Predictive Control for Legged Systems" at ICRA 2024, Yokohama, Japan
Abstract
The problem of designing complex robotic hardware using numerical optimization has gotten significant attention in recent years. However, ensuring robustness, which is essential to guarantee the practical applicability of the designed solutions, remains an ongoing challenge in co-design. This problem becomes particularly pronounced when dealing with inherently unstable systems, such as legged robots. In this extended abstract, we want to reason about these challenges and investigate possible solutions to tackle such a problem based on current trends in the community.
Gabriele Fadini
Workshop "Co-design in Robotics: Theory, Practice, and Challenges"
ICRA 2024, Yokohama, Japan
Abstract
Co-design is a valuable method for selecting optimized robots for a given task. In past research, this approach has been used to enhance the performance of legged robots. In our latest work, the conception of a quadruped prototype was driven by this approach, in order to achieve energy efficiency in cyclic tasks. The optimization of hardware and control features a complete electro-mechanical model that factors in actuator losses and mechanical design considerations. While this co-design formulation is versatile enough to cope with various tasks, the optimization-to-real gap still needs to be properly assessed and possibly reduced in some specific scenarios, hinting to some of the biggest open challenges in the practical application of co-design.
Presentations