Abstract:On-robot Reinforcement Learning is a promising approach to train embodiment-aware policies for legged robots. However, the computational constraints of real-time learning on robots pose a significant challenge. We present a framework for efficiently learning quadruped locomotion in just 8 minutes of raw real-time training utilizing the sample efficiency and minimal computational overhead of the new off-policy algorithm CrossQ. We investigate two control architectures: Predicting joint target positions for agile, high-speed locomotion and Central Pattern Generators for stable, natural gaits. While prior work focused on learning simple forward gaits, our framework extends on-robot learning to omnidirectional locomotion. We demonstrate the robustness of our approach in different indoor and outdoor environments.
Abstract:Robot foundation models hold the potential for deployment across diverse environments, from industrial applications to household tasks. While current research focuses primarily on the policies' generalization capabilities across a variety of tasks, it fails to address safety, a critical requirement for deployment on real-world systems. In this paper, we introduce a safety layer designed to constrain the action space of any generalist policy appropriately. Our approach uses ATACOM, a safe reinforcement learning algorithm that creates a safe action space and, therefore, ensures safe state transitions. By extending ATACOM to generalist policies, our method facilitates their deployment in safety-critical scenarios without requiring any specific safety fine-tuning. We demonstrate the effectiveness of this safety layer in an air hockey environment, where it prevents a puck-hitting agent from colliding with its surroundings, a failure observed in generalist policies.
Abstract:Prior flow matching methods in robotics have primarily learned velocity fields to morph one distribution of trajectories into another. In this work, we extend flow matching to capture second-order trajectory dynamics, incorporating acceleration effects either explicitly in the model or implicitly through the learning objective. Unlike diffusion models, which rely on a noisy forward process and iterative denoising steps, flow matching trains a continuous transformation (flow) that directly maps a simple prior distribution to the target trajectory distribution without any denoising procedure. By modeling trajectories with second-order dynamics, our approach ensures that generated robot motions are smooth and physically executable, avoiding the jerky or dynamically infeasible trajectories that first-order models might produce. We empirically demonstrate that this second-order conditional flow matching yields superior performance on motion planning benchmarks, achieving smoother trajectories and higher success rates than baseline planners. These findings highlight the advantage of learning acceleration-aware motion fields, as our method outperforms existing motion planning methods in terms of trajectory quality and planning success.
Abstract:Recent works have successfully demonstrated that sparse deep reinforcement learning agents can be competitive against their dense counterparts. This opens up opportunities for reinforcement learning applications in fields where inference time and memory requirements are cost-sensitive or limited by hardware. Until now, dense-to-sparse methods have relied on hand-designed sparsity schedules that are not synchronized with the agent's learning pace. Crucially, the final sparsity level is chosen as a hyperparameter, which requires careful tuning as setting it too high might lead to poor performances. In this work, we address these shortcomings by crafting a dense-to-sparse algorithm that we name Eau De $Q$-Network (EauDeQN). To increase sparsity at the agent's learning pace, we consider multiple online networks with different sparsity levels, where each online network is trained from a shared target network. At each target update, the online network with the smallest loss is chosen as the next target network, while the other networks are replaced by a pruned version of the chosen network. We evaluate the proposed approach on the Atari $2600$ benchmark and the MuJoCo physics simulator, showing that EauDeQN reaches high sparsity levels while keeping performances high.
Abstract:We introduce a scaling strategy for Explicit Policy-Conditioned Value Functions (EPVFs) that significantly improves performance on challenging continuous-control tasks. EPVFs learn a value function V({\theta}) that is explicitly conditioned on the policy parameters, enabling direct gradient-based updates to the parameters of any policy. However, EPVFs at scale struggle with unrestricted parameter growth and efficient exploration in the policy parameter space. To address these issues, we utilize massive parallelization with GPU-based simulators, big batch sizes, weight clipping and scaled peturbations. Our results show that EPVFs can be scaled to solve complex tasks, such as a custom Ant environment, and can compete with state-of-the-art Deep Reinforcement Learning (DRL) baselines like Proximal Policy Optimization (PPO) and Soft Actor-Critic (SAC). We further explore action-based policy parameter representations from previous work and specialized neural network architectures to efficiently handle weight-space features, which have not been used in the context of DRL before.
Abstract:Reinforcement learning has achieved significant milestones, but sample efficiency remains a bottleneck for real-world applications. Recently, CrossQ has demonstrated state-of-the-art sample efficiency with a low update-to-data (UTD) ratio of 1. In this work, we explore CrossQ's scaling behavior with higher UTD ratios. We identify challenges in the training dynamics, which are emphasized by higher UTD ratios. To address these, we integrate weight normalization into the CrossQ framework, a solution that stabilizes training, has been shown to prevent potential loss of plasticity and keeps the effective learning rate constant. Our proposed approach reliably scales with increasing UTD ratios, achieving competitive performance across 25 challenging continuous control tasks on the DeepMind Control Suite and Myosuite benchmarks, notably the complex dog and humanoid environments. This work eliminates the need for drastic interventions, such as network resets, and offers a simple yet robust pathway for improving sample efficiency and scalability in model-free reinforcement learning.
Abstract:Maximum entropy reinforcement learning (MaxEnt-RL) has become the standard approach to RL due to its beneficial exploration properties. Traditionally, policies are parameterized using Gaussian distributions, which significantly limits their representational capacity. Diffusion-based policies offer a more expressive alternative, yet integrating them into MaxEnt-RL poses challenges--primarily due to the intractability of computing their marginal entropy. To overcome this, we propose Diffusion-Based Maximum Entropy RL (DIME). DIME leverages recent advances in approximate inference with diffusion models to derive a lower bound on the maximum entropy objective. Additionally, we propose a policy iteration scheme that provably converges to the optimal diffusion policy. Our method enables the use of expressive diffusion-based policies while retaining the principled exploration benefits of MaxEnt-RL, significantly outperforming other diffusion-based methods on challenging high-dimensional control benchmarks. It is also competitive with state-of-the-art non-diffusion based RL methods while requiring fewer algorithmic design choices and smaller update-to-data ratios, reducing computational complexity.
Abstract:In recent years, nonlinear dynamic system identification using artificial neural networks has garnered attention due to its manifold potential applications in virtually all branches of science and engineering. However, purely data-driven approaches often struggle with extrapolation and may yield physically implausible forecasts. Furthermore, the learned dynamics can exhibit instabilities, making it difficult to apply such models safely and robustly. This article proposes stable port-Hamiltonian neural networks, a machine learning architecture that incorporates the physical biases of energy conservation or dissipation while guaranteeing global Lyapunov stability of the learned dynamics. Evaluations with illustrative examples and real-world measurement data demonstrate the model's ability to generalize from sparse data, outperforming purely data-driven approaches and avoiding instability issues. In addition, the model's potential for data-driven surrogate modeling is highlighted in application to multi-physics simulation data.
Abstract:This paper introduces a new imitation learning framework based on energy-based generative models capable of learning complex, physics-dependent, robot motion policies through state-only expert motion trajectories. Our algorithm, called Noise-conditioned Energy-based Annealed Rewards (NEAR), constructs several perturbed versions of the expert's motion data distribution and learns smooth, and well-defined representations of the data distribution's energy function using denoising score matching. We propose to use these learnt energy functions as reward functions to learn imitation policies via reinforcement learning. We also present a strategy to gradually switch between the learnt energy functions, ensuring that the learnt rewards are always well-defined in the manifold of policy-generated samples. We evaluate our algorithm on complex humanoid tasks such as locomotion and martial arts and compare it with state-only adversarial imitation learning algorithms like Adversarial Motion Priors (AMP). Our framework sidesteps the optimisation challenges of adversarial imitation learning techniques and produces results comparable to AMP in several quantitative metrics across multiple imitation settings.
Abstract:Diffusion-based visuomotor policies excel at learning complex robotic tasks by effectively combining visual data with high-dimensional, multi-modal action distributions. However, diffusion models often suffer from slow inference due to costly denoising processes or require complex sequential training arising from recent distilling approaches. This paper introduces Riemannian Flow Matching Policy (RFMP), a model that inherits the easy training and fast inference capabilities of flow matching (FM). Moreover, RFMP inherently incorporates geometric constraints commonly found in realistic robotic applications, as the robot state resides on a Riemannian manifold. To enhance the robustness of RFMP, we propose Stable RFMP (SRFMP), which leverages LaSalle's invariance principle to equip the dynamics of FM with stability to the support of a target Riemannian distribution. Rigorous evaluation on eight simulated and real-world tasks show that RFMP successfully learns and synthesizes complex sensorimotor policies on Euclidean and Riemannian spaces with efficient training and inference phases, outperforming Diffusion Policies while remaining competitive with Consistency Policies.