Markus Wulfmeier

Tuomas Haarnoja, Ben Moran, Guy Lever et al. · deepmind

We investigate whether Deep Reinforcement Learning (Deep RL) is able to synthesize sophisticated and safe movement skills for a low-cost, miniature humanoid robot that can be composed into complex behavioral strategies in dynamic environments. We used Deep RL to train a humanoid robot with 20 actuated joints to play a simplified one-versus-one (1v1) soccer game. The resulting agent exhibits robust and dynamic movement skills such as rapid fall recovery, walking, turning, kicking and more; and it transitions between them in a smooth, stable, and efficient manner. The agent's locomotion and tactical behavior adapts to specific game contexts in a way that would be impractical to manually design. The agent also developed a basic strategic understanding of the game, and learned, for instance, to anticipate ball movements and to block opponent shots. Our agent was trained in simulation and transferred to real robots zero-shot. We found that a combination of sufficiently high-frequency control, targeted dynamics randomization, and perturbations during training in simulation enabled good-quality transfer. Although the robots are inherently fragile, basic regularization of the behavior during training led the robots to learn safe and effective movements while still performing in a dynamic and agile way -- well beyond what is intuitively expected from the robot. Indeed, in experiments, they walked 181% faster, turned 302% faster, took 63% less time to get up, and kicked a ball 34% faster than a scripted baseline, while efficiently combining the skills to achieve the longer term objectives.

Norman Di Palo, Arunkumar Byravan, Leonard Hasenclever et al. · deepmind

Language Models and Vision Language Models have recently demonstrated unprecedented capabilities in terms of understanding human intentions, reasoning, scene understanding, and planning-like behaviour, in text form, among many others. In this work, we investigate how to embed and leverage such abilities in Reinforcement Learning (RL) agents. We design a framework that uses language as the core reasoning tool, exploring how this enables an agent to tackle a series of fundamental RL challenges, such as efficient exploration, reusing experience data, scheduling skills, and learning from observations, which traditionally require separate, vertically designed algorithms. We test our method on a sparse-reward simulated robotic manipulation environment, where a robot needs to stack a set of objects. We demonstrate substantial performance improvements over baselines in exploration efficiency and ability to reuse data from offline datasets, and illustrate how to reuse learned skills to solve novel tasks or imitate videos of human experts.

Tim Seyde, Peter Werner, Wilko Schwarting et al. · deepmind

While there has been substantial success for solving continuous control with actor-critic methods, simpler critic-only methods such as Q-learning find limited application in the associated high-dimensional action spaces. However, most actor-critic methods come at the cost of added complexity: heuristics for stabilisation, compute requirements and wider hyperparameter search spaces. We show that a simple modification of deep Q-learning largely alleviates these issues. By combining bang-bang action discretization with value decomposition, framing single-agent control as cooperative multi-agent reinforcement learning (MARL), this simple critic-only approach matches performance of state-of-the-art continuous actor-critic methods when learning from features or pixels. We extend classical bandit examples from cooperative MARL to provide intuition for how decoupled critics leverage state information to coordinate joint optimization, and demonstrate surprisingly strong performance across a variety of continuous control tasks.

Sasha Salter, Markus Wulfmeier, Dhruva Tirumala et al. · deepmind

The ability to discover useful behaviours from past experience and transfer them to new tasks is considered a core component of natural embodied intelligence. Inspired by neuroscience, discovering behaviours that switch at bottleneck states have been long sought after for inducing plans of minimum description length across tasks. Prior approaches have either only supported online, on-policy, bottleneck state discovery, limiting sample-efficiency, or discrete state-action domains, restricting applicability. To address this, we introduce Model-Based Offline Options (MO2), an offline hindsight framework supporting sample-efficient bottleneck option discovery over continuous state-action spaces. Once bottleneck options are learnt offline over source domains, they are transferred online to improve exploration and value estimation on the transfer domain. Our experiments show that on complex long-horizon continuous control tasks with sparse, delayed rewards, MO2's properties are essential and lead to performance exceeding recent option learning methods. Additional ablations further demonstrate the impact on option predictability and credit assignment.

Dhruva Tirumala, Thomas Lampe, Jose Enrique Chen et al. · deepmind

Replaying data is a principal mechanism underlying the stability and data efficiency of off-policy reinforcement learning (RL). We present an effective yet simple framework to extend the use of replays across multiple experiments, minimally adapting the RL workflow for sizeable improvements in controller performance and research iteration times. At its core, Replay Across Experiments (RaE) involves reusing experience from previous experiments to improve exploration and bootstrap learning while reducing required changes to a minimum in comparison to prior work. We empirically show benefits across a number of RL algorithms and challenging control domains spanning both locomotion and manipulation, including hard exploration tasks from egocentric vision. Through comprehensive ablations, we demonstrate robustness to the quality and amount of data available and various hyperparameter choices. Finally, we discuss how our approach can be applied more broadly across research life cycles and can increase resilience by reloading data across random seeds or hyperparameter variations.

Giulia Vezzani, Dhruva Tirumala, Markus Wulfmeier et al. · deepmind

The ability to effectively reuse prior knowledge is a key requirement when building general and flexible Reinforcement Learning (RL) agents. Skill reuse is one of the most common approaches, but current methods have considerable limitations.For example, fine-tuning an existing policy frequently fails, as the policy can degrade rapidly early in training. In a similar vein, distillation of expert behavior can lead to poor results when given sub-optimal experts. We compare several common approaches for skill transfer on multiple domains including changes in task and system dynamics. We identify how existing methods can fail and introduce an alternative approach to mitigate these problems. Our approach learns to sequence existing temporally-extended skills for exploration but learns the final policy directly from the raw experience. This conceptual split enables rapid adaptation and thus efficient data collection but without constraining the final solution.It significantly outperforms many classical methods across a suite of evaluation tasks and we use a broad set of ablations to highlight the importance of differentc omponents of our method.

Nico Gürtler, Felix Widmaier, Cansu Sancaktar et al. · deepmind

Experimentation on real robots is demanding in terms of time and costs. For this reason, a large part of the reinforcement learning (RL) community uses simulators to develop and benchmark algorithms. However, insights gained in simulation do not necessarily translate to real robots, in particular for tasks involving complex interactions with the environment. The Real Robot Challenge 2022 therefore served as a bridge between the RL and robotics communities by allowing participants to experiment remotely with a real robot - as easily as in simulation. In the last years, offline reinforcement learning has matured into a promising paradigm for learning from pre-collected datasets, alleviating the reliance on expensive online interactions. We therefore asked the participants to learn two dexterous manipulation tasks involving pushing, grasping, and in-hand orientation from provided real-robot datasets. An extensive software documentation and an initial stage based on a simulation of the real set-up made the competition particularly accessible. By giving each team plenty of access budget to evaluate their offline-learned policies on a cluster of seven identical real TriFinger platforms, we organized an exciting competition for machine learners and roboticists alike. In this work we state the rules of the competition, present the methods used by the winning teams and compare their results with a benchmark of state-of-the-art offline RL algorithms on the challenge datasets.

Wenxuan Zhou, Steven Bohez, Jan Humplik et al. · deepmind

Robots will experience non-stationary environment dynamics throughout their lifetime: the robot dynamics can change due to wear and tear, or its surroundings may change over time. Eventually, the robots should perform well in all of the environment variations it has encountered. At the same time, it should still be able to learn fast in a new environment. We identify two challenges in Reinforcement Learning (RL) under such a lifelong learning setting with off-policy data: first, existing off-policy algorithms struggle with the trade-off between being conservative to maintain good performance in the old environment and learning efficiently in the new environment, despite keeping all the data in the replay buffer. We propose the Offline Distillation Pipeline to break this trade-off by separating the training procedure into an online interaction phase and an offline distillation phase.Second, we find that training with the imbalanced off-policy data from multiple environments across the lifetime creates a significant performance drop. We identify that this performance drop is caused by the combination of the imbalanced quality and size among the datasets which exacerbate the extrapolation error of the Q-function. During the distillation phase, we apply a simple fix to the issue by keeping the policy closer to the behavior policy that generated the data. In the experiments, we demonstrate these two challenges and the proposed solutions with a simulated bipedal robot walk-ing task across various environment changes. We show that the Offline Distillation Pipeline achieves better performance across all the encountered environments without affecting data collection. We also provide a comprehensive empirical study to support our hypothesis on the data imbalance issue.

Markus Wulfmeier, Michael Bloesch, Nino Vieillard et al. · deepmind

The majority of language model training builds on imitation learning. It covers pretraining, supervised fine-tuning, and affects the starting conditions for reinforcement learning from human feedback (RLHF). The simplicity and scalability of maximum likelihood estimation (MLE) for next token prediction led to its role as predominant paradigm. However, the broader field of imitation learning can more effectively utilize the sequential structure underlying autoregressive generation. We focus on investigating the inverse reinforcement learning (IRL) perspective to imitation, extracting rewards and directly optimizing sequences instead of individual token likelihoods and evaluate its benefits for fine-tuning large language models. We provide a new angle, reformulating inverse soft-Q-learning as a temporal difference regularized extension of MLE. This creates a principled connection between MLE and IRL and allows trading off added complexity with increased performance and diversity of generations in the supervised fine-tuning (SFT) setting. We find clear advantages for IRL-based imitation, in particular for retaining diversity while maximizing task performance, rendering IRL a strong alternative on fixed SFT datasets even without online data generation. Our analysis of IRL-extracted reward functions further indicates benefits for more robust reward functions via tighter integration of supervised and preference-based LLM post-training.

Cristina Pinneri, Sarah Bechtle, Markus Wulfmeier et al. · deepmind

We present a novel approach to address the challenge of generalization in offline reinforcement learning (RL), where the agent learns from a fixed dataset without any additional interaction with the environment. Specifically, we aim to improve the agent's ability to generalize to out-of-distribution goals. To achieve this, we propose to learn a dynamics model and check if it is equivariant with respect to a fixed type of transformation, namely translations in the state space. We then use an entropy regularizer to increase the equivariant set and augment the dataset with the resulting transformed samples. Finally, we learn a new policy offline based on the augmented dataset, with an off-the-shelf offline RL algorithm. Our experimental results demonstrate that our approach can greatly improve the test performance of the policy on the considered environments.

Yarden As, Dhruva Tirumala, René Zurbrügg et al. · deepmind

We investigate what specific design choices enable successful online reinforcement learning (RL) on physical robots. Across 100 real-world training runs on three distinct robotic platforms, we systematically ablate algorithmic, systems, and experimental decisions that are typically left implicit in prior work. We find that some widely used defaults can be harmful, while a set of robust, readily adopted design choices within standard RL practice yield stable learning across tasks and hardware. These results provide the first large-sample empirical study of such design choices, enabling practitioners to deploy online RL with lower engineering effort.

Ismail Geles, Leonard Bauersfeld, Markus Wulfmeier et al.

Autonomous systems have achieved superhuman performance in isolation or simulation, yet they remain brittle in shared, dynamic real-world spaces. This failure stems from the dominant single-agent paradigm for physical applications, where other actors are ignored or treated as environmental noise, preventing effective coordination. Here we show that multi-agent reinforcement learning provides the essential safety scaffolding required for real-world interaction. Using high-speed quadrotor racing as a high-stakes testbed, we train agents to navigate complex aerodynamic interactions and strategic maneuvering with a variable number of racers. Through league-based self-play, agents evolve sophisticated anticipatory behaviors, including proactive collision avoidance, overtaking, and handling multi-agent physical interactions, including aerodynamic downwash. Our agents outperform a champion-level human pilot in multi-player races at speeds exceeding 22 m/s, while simultaneously reducing collision rates by 50 % compared to state-of-the-art single-agent baselines. Crucially, training with diverse artificial agents enables zero-shot generalization to safer human interaction. These results suggest that the path to robust robotic co-existence lies not in isolated safety constraints, but in the rigorous demands of multi-agent interaction. Multimedia materials are available at: https://rpg.ifi.uzh.ch/marl

Sonia K. Murthy, Rosie Zhao, Jennifer Hu et al. · deepmind

Value trade-offs are an integral part of human decision-making and language use, however, current tools for interpreting such dynamic and multi-faceted notions of values in LLMs are limited. In cognitive science, so-called "cognitive models" provide formal accounts of such trade-offs in humans, by modeling the weighting of a speaker's competing utility functions in choosing an action or utterance. Here we use a leading cognitive model of polite speech to systematically evaluate value trade-offs in two encompassing model settings: degrees of reasoning "effort" in frontier black-box models, and RL post-training dynamics of open-source models. Our results highlight patterns of higher informational utility than social utility in reasoning models' default behavior, and demonstrate that these patterns shift in predictable ways when models are prompted to prioritize certain goals over others. Our findings from LLMs' training dynamics suggest large shifts in utility values early on in training with persistent effects of the choice of base model and pretraining data, compared to feedback dataset or alignment method. Our framework offers a flexible tool for probing value trade-offs across diverse model types, providing insights for generating hypotheses about other social behaviors such as sycophancy and for shaping training regimes that better control trade-offs between values during model development.

Dhruva Tirumala, Markus Wulfmeier, Ben Moran et al. · deepmind

We apply multi-agent deep reinforcement learning (RL) to train end-to-end robot soccer policies with fully onboard computation and sensing via egocentric RGB vision. This setting reflects many challenges of real-world robotics, including active perception, agile full-body control, and long-horizon planning in a dynamic, partially-observable, multi-agent domain. We rely on large-scale, simulation-based data generation to obtain complex behaviors from egocentric vision which can be successfully transferred to physical robots using low-cost sensors. To achieve adequate visual realism, our simulation combines rigid-body physics with learned, realistic rendering via multiple Neural Radiance Fields (NeRFs). We combine teacher-based multi-agent RL and cross-experiment data reuse to enable the discovery of sophisticated soccer strategies. We analyze active-perception behaviors including object tracking and ball seeking that emerge when simply optimizing perception-agnostic soccer play. The agents display equivalent levels of performance and agility as policies with access to privileged, ground-truth state. To our knowledge, this paper constitutes a first demonstration of end-to-end training for multi-agent robot soccer, mapping raw pixel observations to joint-level actions, that can be deployed in the real world. Videos of the game-play and analyses can be seen on our website https://sites.google.com/view/vision-soccer .

Thomas Schmied, Jörg Bornschein, Jordi Grau-Moya et al. · deepmind

The success of Large Language Models (LLMs) has sparked interest in various agentic applications. A key hypothesis is that LLMs, leveraging common sense and Chain-of-Thought (CoT) reasoning, can effectively explore and efficiently solve complex domains. However, LLM agents have been found to suffer from sub-optimal exploration and the knowing-doing gap, the inability to effectively act on knowledge present in the model. In this work, we systematically study why LLMs perform sub-optimally in decision-making scenarios. In particular, we closely examine three prevalent failure modes: greediness, frequency bias, and the knowing-doing gap. We propose mitigation of these shortcomings by fine-tuning via Reinforcement Learning (RL) on self-generated CoT rationales. Our experiments across multi-armed bandits, contextual bandits, and Tic-tac-toe, demonstrate that RL fine-tuning enhances the decision-making abilities of LLMs by increasing exploration and narrowing the knowing-doing gap. Finally, we study both classic exploration mechanisms, such as $ε$-greedy, and LLM-specific approaches, such as self-correction and self-consistency, to enable more effective fine-tuning of LLMs for decision-making.

Markus Wulfmeier, Arunkumar Byravan, Sarah Bechtle et al. · deepmind

Contemporary artificial intelligence systems exhibit rapidly growing abilities accompanied by the growth of required resources, expansive datasets and corresponding investments into computing infrastructure. Although earlier successes predominantly focus on constrained settings, recent strides in fundamental research and applications aspire to create increasingly general systems. This evolving landscape presents a dual panorama of opportunities and challenges in refining the generalisation and transfer of knowledge - the extraction from existing sources and adaptation as a comprehensive foundation for tackling new problems. Within the domain of reinforcement learning (RL), the representation of knowledge manifests through various modalities, including dynamics and reward models, value functions, policies, and the original data. This taxonomy systematically targets these modalities and frames its discussion based on their inherent properties and alignment with different objectives and mechanisms for transfer. Where possible, we aim to provide coarse guidance delineating approaches which address requirements such as limiting environment interactions, maximising computational efficiency, and enhancing generalisation across varying axes of change. Finally, we analyse reasons contributing to the prevalence or scarcity of specific forms of transfer, the inherent potential behind pushing these frontiers, and underscore the significance of transitioning from designed to learned transfer.

Tim Seyde, Peter Werner, Wilko Schwarting et al. · deepmind

Recent reinforcement learning approaches have shown surprisingly strong capabilities of bang-bang policies for solving continuous control benchmarks. The underlying coarse action space discretizations often yield favourable exploration characteristics while final performance does not visibly suffer in the absence of action penalization in line with optimal control theory. In robotics applications, smooth control signals are commonly preferred to reduce system wear and energy efficiency, but action costs can be detrimental to exploration during early training. In this work, we aim to bridge this performance gap by growing discrete action spaces from coarse to fine control resolution, taking advantage of recent results in decoupled Q-learning to scale our approach to high-dimensional action spaces up to dim(A) = 38. Our work indicates that an adaptive control resolution in combination with value decomposition yields simple critic-only algorithms that yield surprisingly strong performance on continuous control tasks.

Jonas Buchli, Brendan Tracey, Tomislav Andric et al. · deepmind

Improved low-frequency sensitivity of gravitational wave observatories would unlock study of intermediate-mass black hole mergers, binary black hole eccentricity, and provide early warnings for multi-messenger observations of binary neutron star mergers. Today's mirror stabilization control injects harmful noise, constituting a major obstacle to sensitivity improvements. We eliminated this noise through Deep Loop Shaping, a reinforcement learning method using frequency domain rewards. We proved our methodology on the LIGO Livingston Observatory (LLO). Our controller reduced control noise in the 10--30Hz band by over 30x, and up to 100x in sub-bands surpassing the design goal motivated by the quantum limit. These results highlight the potential of Deep Loop Shaping to improve current and future GW observatories, and more broadly instrumentation and control systems.

Qiyao Wei, Samuel Holt, Jing Yang et al. · deepmind

Peer review, the bedrock of scientific advancement in machine learning (ML), is strained by a crisis of scale. Exponential growth in manuscript submissions to premier ML venues such as NeurIPS, ICML, and ICLR is outpacing the finite capacity of qualified reviewers, leading to concerns about review quality, consistency, and reviewer fatigue. This position paper argues that AI-assisted peer review must become an urgent research and infrastructure priority. We advocate for a comprehensive AI-augmented ecosystem, leveraging Large Language Models (LLMs) not as replacements for human judgment, but as sophisticated collaborators for authors, reviewers, and Area Chairs (ACs). We propose specific roles for AI in enhancing factual verification, guiding reviewer performance, assisting authors in quality improvement, and supporting ACs in decision-making. Crucially, we contend that the development of such systems hinges on access to more granular, structured, and ethically-sourced peer review process data. We outline a research agenda, including illustrative experiments, to develop and validate these AI assistants, and discuss significant technical and ethical challenges. We call upon the ML community to proactively build this AI-assisted future, ensuring the continued integrity and scalability of scientific validation, while maintaining high standards of peer review.

Mohak Bhardwaj, Thomas Lampe, Michael Neunert et al. · deepmind

Recent advances in real-world applications of reinforcement learning (RL) have relied on the ability to accurately simulate systems at scale. However, domains such as fluid dynamical systems exhibit complex dynamic phenomena that are hard to simulate at high integration rates, limiting the direct application of modern deep RL algorithms to often expensive or safety critical hardware. In this work, we introduce "Box o Flows", a novel benchtop experimental control system for systematically evaluating RL algorithms in dynamic real-world scenarios. We describe the key components of the Box o Flows, and through a series of experiments demonstrate how state-of-the-art model-free RL algorithms can synthesize a variety of complex behaviors via simple reward specifications. Furthermore, we explore the role of offline RL in data-efficient hypothesis testing by reusing past experiences. We believe that the insights gained from this preliminary study and the availability of systems like the Box o Flows support the way forward for developing systematic RL algorithms that can be generally applied to complex, dynamical systems. Supplementary material and videos of experiments are available at https://sites.google.com/view/box-o-flows/home.

Annie S. Chen, Philemon Brakel, Antonia Bronars et al. · deepmind

Learning-based methods for dexterous manipulation have made notable progress in recent years. However, learned policies often still lack reliability and exhibit limited robustness to important factors of variation. One failure pattern that can be observed across many settings is that policies idle, i.e. they cease to move beyond a small region of states when they reach certain states. This policy idling is often a reflection of the training data. For instance, it can occur when the data contains small actions in areas where the robot needs to perform high-precision motions, e.g., when preparing to grasp an object or object insertion. Prior works have tried to mitigate this phenomenon e.g. by filtering the training data or modifying the control frequency. However, these approaches can negatively impact policy performance in other ways. As an alternative, we investigate how to leverage the detectability of idling behavior to inform exploration and policy improvement. Our approach, Pause-Induced Perturbations (PIP), applies perturbations at detected idling states, thus helping it to escape problematic basins of attraction. On a range of challenging simulated dual-arm tasks, we find that this simple approach can already noticeably improve test-time performance, with no additional supervision or training. Furthermore, since the robot tends to idle at critical points in a movement, we also find that learning from the resulting episodes leads to better iterative policy improvement compared to prior approaches. Our perturbation strategy also leads to a 15-35% improvement in absolute success rate on a real-world insertion task that requires complex multi-finger manipulation.

Michael Bloesch, Markus Wulfmeier, Philemon Brakel et al.

Imitation Learning from Observation (IfO) offers a powerful way to learn behaviors at large-scale: Unlike behavior cloning or offline reinforcement learning, IfO can leverage action-free demonstrations and thus circumvents the need for costly action-labeled demonstrations or reward functions. However, current IfO research focuses on idealized scenarios with mostly bimodal-quality data distributions, restricting the meaningfulness of the results. In contrast, this paper investigates more nuanced distributions and introduces a method to learn from such data, moving closer to a paradigm in which imitation learning can be performed iteratively via self-improvement. Our method adapts RL-based imitation learning to action-free demonstrations, using a value function to transfer information between expert and non-expert data. Through comprehensive evaluation, we delineate the relation between different data distributions and the applicability of algorithms and highlight the limitations of established methods. Our findings provide valuable insights for developing more robust and practical IfO techniques on a path to scalable behaviour learning.

Matt Barnes, Matthew Abueg, Oliver F. Lange et al.

Inverse reinforcement learning (IRL) offers a powerful and general framework for learning humans' latent preferences in route recommendation, yet no approach has successfully addressed planetary-scale problems with hundreds of millions of states and demonstration trajectories. In this paper, we introduce scaling techniques based on graph compression, spatial parallelization, and improved initialization conditions inspired by a connection to eigenvector algorithms. We revisit classic IRL methods in the routing context, and make the key observation that there exists a trade-off between the use of cheap, deterministic planners and expensive yet robust stochastic policies. This insight is leveraged in Receding Horizon Inverse Planning (RHIP), a new generalization of classic IRL algorithms that provides fine-grained control over performance trade-offs via its planning horizon. Our contributions culminate in a policy that achieves a 16-24% improvement in route quality at a global scale, and to the best of our knowledge, represents the largest published study of IRL algorithms in a real-world setting to date. We conclude by conducting an ablation study of key components, presenting negative results from alternative eigenvalue solvers, and identifying opportunities to further improve scalability via IRL-specific batching strategies.

Steven Bohez, Saran Tunyasuvunakool, Philemon Brakel et al.

We investigate the use of prior knowledge of human and animal movement to learn reusable locomotion skills for real legged robots. Our approach builds upon previous work on imitating human or dog Motion Capture (MoCap) data to learn a movement skill module. Once learned, this skill module can be reused for complex downstream tasks. Importantly, due to the prior imposed by the MoCap data, our approach does not require extensive reward engineering to produce sensible and natural looking behavior at the time of reuse. This makes it easy to create well-regularized, task-oriented controllers that are suitable for deployment on real robots. We demonstrate how our skill module can be used for imitation, and train controllable walking and ball dribbling policies for both the ANYmal quadruped and OP3 humanoid. These policies are then deployed on hardware via zero-shot simulation-to-reality transfer. Accompanying videos are available at https://bit.ly/robot-npmp.

Nathan Lambert, Markus Wulfmeier, William Whitney et al.

Offline Reinforcement Learning (ORL) enablesus to separately study the two interlinked processes of reinforcement learning: collecting informative experience and inferring optimal behaviour. The second step has been widely studied in the offline setting, but just as critical to data-efficient RL is the collection of informative data. The task-agnostic setting for data collection, where the task is not known a priori, is of particular interest due to the possibility of collecting a single dataset and using it to solve several downstream tasks as they arise. We investigate this setting via curiosity-based intrinsic motivation, a family of exploration methods which encourage the agent to explore those states or transitions it has not yet learned to model. With Explore2Offline, we propose to evaluate the quality of collected data by transferring the collected data and inferring policies with reward relabelling and standard offline RL algorithms. We evaluate a wide variety of data collection strategies, including a new exploration agent, Intrinsic Model Predictive Control (IMPC), using this scheme and demonstrate their performance on various tasks. We use this decoupled framework to strengthen intuitions about exploration and the data prerequisites for effective offline RL.

Dushyant Rao, Fereshteh Sadeghi, Leonard Hasenclever et al.

For robots operating in the real world, it is desirable to learn reusable behaviours that can effectively be transferred and adapted to numerous tasks and scenarios. We propose an approach to learn abstract motor skills from data using a hierarchical mixture latent variable model. In contrast to existing work, our method exploits a three-level hierarchy of both discrete and continuous latent variables, to capture a set of high-level behaviours while allowing for variance in how they are executed. We demonstrate in manipulation domains that the method can effectively cluster offline data into distinct, executable behaviours, while retaining the flexibility of a continuous latent variable model. The resulting skills can be transferred and fine-tuned on new tasks, unseen objects, and from state to vision-based policies, yielding better sample efficiency and asymptotic performance compared to existing skill- and imitation-based methods. We further analyse how and when the skills are most beneficial: they encourage directed exploration to cover large regions of the state space relevant to the task, making them most effective in challenging sparse-reward settings.

Todor Davchev, Oleg Sushkov, Jean-Baptiste Regli et al.

Complex sequential tasks in continuous-control settings often require agents to successfully traverse a set of "narrow passages" in their state space. Solving such tasks with a sparse reward in a sample-efficient manner poses a challenge to modern reinforcement learning (RL) due to the associated long-horizon nature of the problem and the lack of sufficient positive signal during learning. Various tools have been applied to address this challenge. When available, large sets of demonstrations can guide agent exploration. Hindsight relabelling on the other hand does not require additional sources of information. However, existing strategies explore based on task-agnostic goal distributions, which can render the solution of long-horizon tasks impractical. In this work, we extend hindsight relabelling mechanisms to guide exploration along task-specific distributions implied by a small set of successful demonstrations. We evaluate the approach on four complex, single and dual arm, robotics manipulation tasks against strong suitable baselines. The method requires far fewer demonstrations to solve all tasks and achieves a significantly higher overall performance as task complexity increases. Finally, we investigate the robustness of the proposed solution with respect to the quality of input representations and the number of demonstrations.

Tim Seyde, Igor Gilitschenski, Wilko Schwarting et al.

Reinforcement learning (RL) for continuous control typically employs distributions whose support covers the entire action space. In this work, we investigate the colloquially known phenomenon that trained agents often prefer actions at the boundaries of that space. We draw theoretical connections to the emergence of bang-bang behavior in optimal control, and provide extensive empirical evaluation across a variety of recent RL algorithms. We replace the normal Gaussian by a Bernoulli distribution that solely considers the extremes along each action dimension - a bang-bang controller. Surprisingly, this achieves state-of-the-art performance on several continuous control benchmarks - in contrast to robotic hardware, where energy and maintenance cost affect controller choices. Since exploration, learning,and the final solution are entangled in RL, we provide additional imitation learning experiments to reduce the impact of exploration on our analysis. Finally, we show that our observations generalize to environments that aim to model real-world challenges and evaluate factors to mitigate the emergence of bang-bang solutions. Our findings emphasize challenges for benchmarking continuous control algorithms, particularly in light of potential real-world applications.

Oliver Groth, Markus Wulfmeier, Giulia Vezzani et al.

Curiosity-based reward schemes can present powerful exploration mechanisms which facilitate the discovery of solutions for complex, sparse or long-horizon tasks. However, as the agent learns to reach previously unexplored spaces and the objective adapts to reward new areas, many behaviours emerge only to disappear due to being overwritten by the constantly shifting objective. We argue that merely using curiosity for fast environment exploration or as a bonus reward for a specific task does not harness the full potential of this technique and misses useful skills. Instead, we propose to shift the focus towards retaining the behaviours which emerge during curiosity-based learning. We posit that these self-discovered behaviours serve as valuable skills in an agent's repertoire to solve related tasks. Our experiments demonstrate the continuous shift in behaviour throughout training and the benefits of a simple policy snapshot method to reuse discovered behaviour for transfer tasks.

Siqi Liu, Guy Lever, Zhe Wang et al.

Intelligent behaviour in the physical world exhibits structure at multiple spatial and temporal scales. Although movements are ultimately executed at the level of instantaneous muscle tensions or joint torques, they must be selected to serve goals defined on much longer timescales, and in terms of relations that extend far beyond the body itself, ultimately involving coordination with other agents. Recent research in artificial intelligence has shown the promise of learning-based approaches to the respective problems of complex movement, longer-term planning and multi-agent coordination. However, there is limited research aimed at their integration. We study this problem by training teams of physically simulated humanoid avatars to play football in a realistic virtual environment. We develop a method that combines imitation learning, single- and multi-agent reinforcement learning and population-based training, and makes use of transferable representations of behaviour for decision making at different levels of abstraction. In a sequence of stages, players first learn to control a fully articulated body to perform realistic, human-like movements such as running and turning; they then acquire mid-level football skills such as dribbling and shooting; finally, they develop awareness of others and play as a team, bridging the gap between low-level motor control at a timescale of milliseconds, and coordinated goal-directed behaviour as a team at the timescale of tens of seconds. We investigate the emergence of behaviours at different levels of abstraction, as well as the representations that underlie these behaviours using several analysis techniques, including statistics from real-world sports analytics. Our work constitutes a complete demonstration of integrated decision-making at multiple scales in a physically embodied multi-agent setting. See project video at https://youtu.be/KHMwq9pv7mg.

Markus Wulfmeier, Arunkumar Byravan, Tim Hertweck et al.

Projecting high-dimensional environment observations into lower-dimensional structured representations can considerably improve data-efficiency for reinforcement learning in domains with limited data such as robotics. Can a single generally useful representation be found? In order to answer this question, it is important to understand how the representation will be used by the agent and what properties such a 'good' representation should have. In this paper we systematically evaluate a number of common learnt and hand-engineered representations in the context of three robotics tasks: lifting, stacking and pushing of 3D blocks. The representations are evaluated in two use-cases: as input to the agent, or as a source of auxiliary tasks. Furthermore, the value of each representation is evaluated in terms of three properties: dimensionality, observability and disentanglement. We can significantly improve performance in both use-cases and demonstrate that some representations can perform commensurate to simulator states as agent inputs. Finally, our results challenge common intuitions by demonstrating that: 1) dimensionality strongly matters for task generation, but is negligible for inputs, 2) observability of task-relevant aspects mostly affects the input representation use-case, and 3) disentanglement leads to better auxiliary tasks, but has only limited benefits for input representations. This work serves as a step towards a more systematic understanding of what makes a 'good' representation for control in robotics, enabling practitioners to make more informed choices for developing new learned or hand-engineered representations.

Giulia Vezzani, Michael Neunert, Markus Wulfmeier et al.

Robot manipulation requires a complex set of skills that need to be carefully combined and coordinated to solve a task. Yet, most ReinforcementLearning (RL) approaches in robotics study tasks which actually consist only of a single manipulation skill, such as grasping an object or inserting a pre-grasped object. As a result the skill ('how' to solve the task) but not the actual goal of a complete manipulation ('what' to solve) is specified. In contrast, we study a complex manipulation goal that requires an agent to learn and combine diverse manipulation skills. We propose a challenging, highly under-actuated peg-in-hole task with a free, rotational asymmetrical peg, requiring a broad range of manipulation skills. While correct peg (re-)orientation is a requirement for successful insertion, there is no reward associated with it. Hence an agent needs to understand this pre-condition and learn the skill to fulfil it. The final insertion reward is sparse, allowing freedom in the solution and leading to complex emerging behaviour not envisioned during the task design. We tackle the problem in a multi-task RL framework using Scheduled Auxiliary Control (SAC-X) combined with Regularized Hierarchical Policy Optimization (RHPO) which successfully solves the task in simulation and from scratch on a single robot where data is severely limited.

Roland Hafner, Tim Hertweck, Philipp Klöppner et al.

Modern Reinforcement Learning (RL) algorithms promise to solve difficult motor control problems directly from raw sensory inputs. Their attraction is due in part to the fact that they can represent a general class of methods that allow to learn a solution with a reasonably set reward and minimal prior knowledge, even in situations where it is difficult or expensive for a human expert. For RL to truly make good on this promise, however, we need algorithms and learning setups that can work across a broad range of problems with minimal problem specific adjustments or engineering. In this paper, we study this idea of generality in the locomotion domain. We develop a learning framework that can learn sophisticated locomotion behavior for a wide spectrum of legged robots, such as bipeds, tripeds, quadrupeds and hexapods, including wheeled variants. Our learning framework relies on a data-efficient, off-policy multi-task RL algorithm and a small set of reward functions that are semantically identical across robots. To underline the general applicability of the method, we keep the hyper-parameter settings and reward definitions constant across experiments and rely exclusively on on-board sensing. For nine different types of robots, including a real-world quadruped robot, we demonstrate that the same algorithm can rapidly learn diverse and reusable locomotion skills without any platform specific adjustments or additional instrumentation of the learning setup.

Markus Wulfmeier, Dushyant Rao, Roland Hafner et al.

We introduce Hindsight Off-policy Options (HO2), a data-efficient option learning algorithm. Given any trajectory, HO2 infers likely option choices and backpropagates through the dynamic programming inference procedure to robustly train all policy components off-policy and end-to-end. The approach outperforms existing option learning methods on common benchmarks. To better understand the option framework and disentangle benefits from both temporal and action abstraction, we evaluate ablations with flat policies and mixture policies with comparable optimization. The results highlight the importance of both types of abstraction as well as off-policy training and trust-region constraints, particularly in challenging, simulated 3D robot manipulation tasks from raw pixel inputs. Finally, we intuitively adapt the inference step to investigate the effect of increased temporal abstraction on training with pre-trained options and from scratch.

Tim Hertweck, Martin Riedmiller, Michael Bloesch et al.

Modern reinforcement learning algorithms can learn solutions to increasingly difficult control problems while at the same time reduce the amount of prior knowledge needed for their application. One of the remaining challenges is the definition of reward schemes that appropriately facilitate exploration without biasing the solution in undesirable ways, and that can be implemented on real robotic systems without expensive instrumentation. In this paper we focus on a setting in which goal tasks are defined via simple sparse rewards, and exploration is facilitated via agent-internal auxiliary tasks. We introduce the idea of simple sensor intentions (SSIs) as a generic way to define auxiliary tasks. SSIs reduce the amount of prior knowledge that is required to define suitable rewards. They can further be computed directly from raw sensor streams and thus do not require expensive and possibly brittle state estimation on real systems. We demonstrate that a learning system based on these rewards can solve complex robotic tasks in simulation and in real world settings. In particular, we show that a real robotic arm can learn to grasp and lift and solve a Ball-in-a-Cup task from scratch, when only raw sensor streams are used for both controller input and in the auxiliary reward definition.

Michael Neunert, Abbas Abdolmaleki, Markus Wulfmeier et al.

Many real-world control problems involve both discrete decision variables - such as the choice of control modes, gear switching or digital outputs - as well as continuous decision variables - such as velocity setpoints, control gains or analogue outputs. However, when defining the corresponding optimal control or reinforcement learning problem, it is commonly approximated with fully continuous or fully discrete action spaces. These simplifications aim at tailoring the problem to a particular algorithm or solver which may only support one type of action space. Alternatively, expert heuristics are used to remove discrete actions from an otherwise continuous space. In contrast, we propose to treat hybrid problems in their 'native' form by solving them with hybrid reinforcement learning, which optimizes for discrete and continuous actions simultaneously. In our experiments, we first demonstrate that the proposed approach efficiently solves such natively hybrid reinforcement learning problems. We then show, both in simulation and on robotic hardware, the benefits of removing possibly imperfect expert-designed heuristics. Lastly, hybrid reinforcement learning encourages us to rethink problem definitions. We propose reformulating control problems, e.g. by adding meta actions, to improve exploration or reduce mechanical wear and tear.

Christopher Grimm, Irina Higgins, Andre Barreto et al.

Biological intelligence can learn to solve many diverse tasks in a data efficient manner by re-using basic knowledge and skills from one task to another. Furthermore, many of such skills are acquired without explicit supervision in an intrinsically driven fashion. This is in contrast to the state-of-the-art reinforcement learning agents, which typically start learning each new task from scratch and struggle with knowledge transfer. In this paper we propose a principled way to learn a basis set of policies, which, when recombined through generalised policy improvement, come with guarantees on the coverage of the final task space. In particular, we concentrate on solving goal-based downstream tasks where the execution order of actions is not important. We demonstrate both theoretically and empirically that learning a small number of policies that reach intrinsically specified goal regions in a disentangled latent space can be re-used to quickly achieve a high level of performance on an exponentially larger number of externally specified, often significantly more complex downstream tasks. Our learning pipeline consists of two stages. First, the agent learns to perform intrinsically generated, goal-based tasks in the total absence of environmental rewards. Second, the agent leverages this experience to quickly achieve a high level of performance on numerous diverse externally specified tasks.

Sasha Salter, Dushyant Rao, Markus Wulfmeier et al.

Image-based Reinforcement Learning is known to suffer from poor sample efficiency and generalisation to unseen visuals such as distractors (task-independent aspects of the observation space). Visual domain randomisation encourages transfer by training over visual factors of variation that may be encountered in the target domain. This increases learning complexity, can negatively impact learning rate and performance, and requires knowledge of potential variations during deployment. In this paper, we introduce Attention-Privileged Reinforcement Learning (APRiL) which uses a self-supervised attention mechanism to significantly alleviate these drawbacks: by focusing on task-relevant aspects of the observations, attention provides robustness to distractors as well as significantly increased learning efficiency. APRiL trains two attention-augmented actor-critic agents: one purely based on image observations, available across training and transfer domains; and one with access to privileged information (such as environment states) available only during training. Experience is shared between both agents and their attention mechanisms are aligned. The image-based policy can then be deployed without access to privileged information. We experimentally demonstrate accelerated and more robust learning on a diverse set of domains, leading to improved final performance for environments both within and outside the training distribution.

Markus Wulfmeier, Abbas Abdolmaleki, Roland Hafner et al.

The successful application of general reinforcement learning algorithms to real-world robotics applications is often limited by their high data requirements. We introduce Regularized Hierarchical Policy Optimization (RHPO) to improve data-efficiency for domains with multiple dominant tasks and ultimately reduce required platform time. To this end, we employ compositional inductive biases on multiple levels and corresponding mechanisms for sharing off-policy transition data across low-level controllers and tasks as well as scheduling of tasks. The presented algorithm enables stable and fast learning for complex, real-world domains in the parallel multitask and sequential transfer case. We show that the investigated types of hierarchy enable positive transfer while partially mitigating negative interference and evaluate the benefits of additional incentives for efficient, compositional task solutions in single task domains. Finally, we demonstrate substantial data-efficiency and final performance gains over competitive baselines in a week-long, physical robot stacking experiment.

Markus Wulfmeier

Recent successes in machine learning have led to a shift in the design of autonomous systems, improving performance on existing tasks and rendering new applications possible. Data-focused approaches gain relevance across diverse, intricate applications when developing data collection and curation pipelines becomes more effective than manual behaviour design. The following work aims at increasing the efficiency of this pipeline in two principal ways: by utilising more powerful sources of informative data and by extracting additional information from existing data. In particular, we target three orthogonal fronts: imitation learning, domain adaptation, and transfer from simulation.

Markus Wulfmeier

Due to recent advances - compute, data, models - the role of learning in autonomous systems has expanded significantly, rendering new applications possible for the first time. While some of the most significant benefits are obtained in the perception modules of the software stack, other aspects continue to rely on known manual procedures based on prior knowledge on geometry, dynamics, kinematics etc. Nonetheless, learning gains relevance in these modules when data collection and curation become easier than manual rule design. Building on this coarse and broad survey of current research, the final sections aim to provide insights into future potentials and challenges as well as the necessity of structure in current practical applications.

Fabian B. Fuchs, Oliver Groth, Adam R. Kosiorek et al.

Visually predicting the stability of block towers is a popular task in the domain of intuitive physics. While previous work focusses on prediction accuracy, a one-dimensional performance measure, we provide a broader analysis of the learned physical understanding of the final model and how the learning process can be guided. To this end, we introduce neural stethoscopes as a general purpose framework for quantifying the degree of importance of specific factors of influence in deep neural networks as well as for actively promoting and suppressing information as appropriate. In doing so, we unify concepts from multitask learning as well as training with auxiliary and adversarial losses. We apply neural stethoscopes to analyse the state-of-the-art neural network for stability prediction. We show that the baseline model is susceptible to being misled by incorrect visual cues. This leads to a performance breakdown to the level of random guessing when training on scenarios where visual cues are inversely correlated with stability. Using stethoscopes to promote meaningful feature extraction increases performance from 51% to 90% prediction accuracy. Conversely, training on an easy dataset where visual cues are positively correlated with stability, the baseline model learns a bias leading to poor performance on a harder dataset. Using an adversarial stethoscope, the network is successfully de-biased, leading to a performance increase from 66% to 88%.

Kyriacos Shiarlis, Markus Wulfmeier, Sasha Salter et al.

Many advanced Learning from Demonstration (LfD) methods consider the decomposition of complex, real-world tasks into simpler sub-tasks. By reusing the corresponding sub-policies within and between tasks, they provide training data for each policy from different high-level tasks and compose them to perform novel ones. Existing approaches to modular LfD focus either on learning a single high-level task or depend on domain knowledge and temporal segmentation. In contrast, we propose a weakly supervised, domain-agnostic approach based on task sketches, which include only the sequence of sub-tasks performed in each demonstration. Our approach simultaneously aligns the sketches with the observed demonstrations and learns the required sub-policies. This improves generalisation in comparison to separate optimisation procedures. We evaluate the approach on multiple domains, including a simulated 3D robot arm control task using purely image-based observations. The results show that our approach performs commensurately with fully supervised approaches, while requiring significantly less annotation effort.

Markus Wulfmeier, Alex Bewley, Ingmar Posner

Continuous appearance shifts such as changes in weather and lighting conditions can impact the performance of deployed machine learning models. While unsupervised domain adaptation aims to address this challenge, current approaches do not utilise the continuity of the occurring shifts. In particular, many robotics applications exhibit these conditions and thus facilitate the potential to incrementally adapt a learnt model over minor shifts which integrate to massive differences over time. Our work presents an adversarial approach for lifelong, incremental domain adaptation which benefits from unsupervised alignment to a series of intermediate domains which successively diverge from the labelled source domain. We empirically demonstrate that our incremental approach improves handling of large appearance changes, e.g. day to night, on a traversable-path segmentation task compared with a direct, single alignment step approach. Furthermore, by approximating the feature distribution for the source domain with a generative adversarial network, the deployment module can be rendered fully independent of retaining potentially large amounts of the related source training data for only a minor reduction in performance.

Markus Wulfmeier, Ingmar Posner, Pieter Abbeel

Training robots for operation in the real world is a complex, time consuming and potentially expensive task. Despite significant success of reinforcement learning in games and simulations, research in real robot applications has not been able to match similar progress. While sample complexity can be reduced by training policies in simulation, such policies can perform sub-optimally on the real platform given imperfect calibration of model dynamics. We present an approach -- supplemental to fine tuning on the real robot -- to further benefit from parallel access to a simulator during training and reduce sample requirements on the real robot. The developed approach harnesses auxiliary rewards to guide the exploration for the real world agent based on the proficiency of the agent in simulation and vice versa. In this context, we demonstrate empirically that the reciprocal alignment for both agents provides further benefit as the agent in simulation can adjust to optimize its behaviour for states commonly visited by the real-world agent.

Carlos Florensa, David Held, Markus Wulfmeier et al.

Many relevant tasks require an agent to reach a certain state, or to manipulate objects into a desired configuration. For example, we might want a robot to align and assemble a gear onto an axle or insert and turn a key in a lock. These goal-oriented tasks present a considerable challenge for reinforcement learning, since their natural reward function is sparse and prohibitive amounts of exploration are required to reach the goal and receive some learning signal. Past approaches tackle these problems by exploiting expert demonstrations or by manually designing a task-specific reward shaping function to guide the learning agent. Instead, we propose a method to learn these tasks without requiring any prior knowledge other than obtaining a single state in which the task is achieved. The robot is trained in reverse, gradually learning to reach the goal from a set of start states increasingly far from the goal. Our method automatically generates a curriculum of start states that adapts to the agent's performance, leading to efficient training on goal-oriented tasks. We demonstrate our approach on difficult simulated navigation and fine-grained manipulation problems, not solvable by state-of-the-art reinforcement learning methods.

Markus Wulfmeier, Alex Bewley, Ingmar Posner

Appearance changes due to weather and seasonal conditions represent a strong impediment to the robust implementation of machine learning systems in outdoor robotics. While supervised learning optimises a model for the training domain, it will deliver degraded performance in application domains that underlie distributional shifts caused by these changes. Traditionally, this problem has been addressed via the collection of labelled data in multiple domains or by imposing priors on the type of shift between both domains. We frame the problem in the context of unsupervised domain adaptation and develop a framework for applying adversarial techniques to adapt popular, state-of-the-art network architectures with the additional objective to align features across domains. Moreover, as adversarial training is notoriously unstable, we first perform an extensive ablation study, adapting many techniques known to stabilise generative adversarial networks, and evaluate on a surrogate classification task with the same appearance change. The distilled insights are applied to the problem of free-space segmentation for motion planning in autonomous driving.

Markus Wulfmeier, Dushyant Rao, Ingmar Posner

Recent advances have shown the capability of Fully Convolutional Neural Networks (FCN) to model cost functions for motion planning in the context of learning driving preferences purely based on demonstration data from human drivers. While pure learning from demonstrations in the framework of Inverse Reinforcement Learning (IRL) is a promising approach, we can benefit from well informed human priors and incorporate them into the learning process. Our work achieves this by pretraining a model to regress to a manual cost function and refining it based on Maximum Entropy Deep Inverse Reinforcement Learning. When injecting prior knowledge as pretraining for the network, we achieve higher robustness, more visually distinct obstacle boundaries, and the ability to capture instances of obstacles that elude models that purely learn from demonstration data. Furthermore, by exploiting these human priors, the resulting model can more accurately handle corner cases that are scarcely seen in the demonstration data, such as stairs, slopes, and underpasses.

Markus Wulfmeier, Dominic Zeng Wang, Ingmar Posner

In this work, we present an approach to learn cost maps for driving in complex urban environments from a very large number of demonstrations of driving behaviour by human experts. The learned cost maps are constructed directly from raw sensor measurements, bypassing the effort of manually designing cost maps as well as features. When deploying the learned cost maps, the trajectories generated not only replicate human-like driving behaviour but are also demonstrably robust against systematic errors in putative robot configuration. To achieve this we deploy a Maximum Entropy based, non-linear IRL framework which uses Fully Convolutional Neural Networks (FCNs) to represent the cost model underlying expert driving behaviour. Using a deep, parametric approach enables us to scale efficiently to large datasets and complex behaviours by being run-time independent of dataset extent during deployment. We demonstrate the scalability and the performance of the proposed approach on an ambitious dataset collected over the course of one year including more than 25k demonstration trajectories extracted from over 120km of driving around pedestrianised areas in the city of Milton Keynes, UK. We evaluate the resulting cost representations by showing the advantages over a carefully manually designed cost map and, in addition, demonstrate its robustness to systematic errors by learning precise cost-maps even in the presence of system calibration perturbations.

Markus Wulfmeier, Peter Ondruska, Ingmar Posner

This paper presents a general framework for exploiting the representational capacity of neural networks to approximate complex, nonlinear reward functions in the context of solving the inverse reinforcement learning (IRL) problem. We show in this context that the Maximum Entropy paradigm for IRL lends itself naturally to the efficient training of deep architectures. At test time, the approach leads to a computational complexity independent of the number of demonstrations, which makes it especially well-suited for applications in life-long learning scenarios. Our approach achieves performance commensurate to the state-of-the-art on existing benchmarks while exceeding on an alternative benchmark based on highly varying reward structures. Finally, we extend the basic architecture - which is equivalent to a simplified subclass of Fully Convolutional Neural Networks (FCNNs) with width one - to include larger convolutions in order to eliminate dependency on precomputed spatial features and work on raw input representations.