Roberto Calandra

Haozhi Qi, Ashish Kumar, Roberto Calandra et al. · berkeley

Generalized in-hand manipulation has long been an unsolved challenge of robotics. As a small step towards this grand goal, we demonstrate how to design and learn a simple adaptive controller to achieve in-hand object rotation using only fingertips. The controller is trained entirely in simulation on only cylindrical objects, which then - without any fine-tuning - can be directly deployed to a real robot hand to rotate dozens of objects with diverse sizes, shapes, and weights over the z-axis. This is achieved via rapid online adaptation of the controller to the object properties using only proprioception history. Furthermore, natural and stable finger gaits automatically emerge from training the control policy via reinforcement learning. Code and more videos are available at https://haozhi.io/hora

Haozhi Qi, Brent Yi, Sudharshan Suresh et al. · berkeley

We introduce RotateIt, a system that enables fingertip-based object rotation along multiple axes by leveraging multimodal sensory inputs. Our system is trained in simulation, where it has access to ground-truth object shapes and physical properties. Then we distill it to operate on realistic yet noisy simulated visuotactile and proprioceptive sensory inputs. These multimodal inputs are fused via a visuotactile transformer, enabling online inference of object shapes and physical properties during deployment. We show significant performance improvements over prior methods and the importance of visual and tactile sensing.

Nathan Lambert, Kristofer Pister, Roberto Calandra

Accurately predicting the consequences of agents' actions is a key prerequisite for planning in robotic control. Model-based reinforcement learning (MBRL) is one paradigm which relies on the iterative learning and prediction of state-action transitions to solve a task. Deep MBRL has become a popular candidate, using a neural network to learn a dynamics model that predicts with each pass from high-dimensional states to actions. These "one-step" predictions are known to become inaccurate over longer horizons of composed prediction - called the compounding error problem. Given the prevalence of the compounding error problem in MBRL and related fields of data-driven control, we set out to understand the properties of and conditions causing these long-horizon errors. In this paper, we explore the effects of subcomponents of a control problem on long term prediction error: including choosing a system, collecting data, and training a model. These detailed quantitative studies on simulated and real-world data show that the underlying dynamics of a system are the strongest factor determining the shape and magnitude of prediction error. Given a clearer understanding of compounding prediction error, researchers can implement new types of models beyond "one-step" that are more useful for control.

Niklas Funk, Changqi Chen, Tim Schneider et al.

The field of robotic manipulation has advanced significantly in recent years. At the sensing level, several novel tactile sensors have been developed, capable of providing accurate contact information. On a methodological level, learning from demonstrations has proven an efficient paradigm to obtain performant robotic manipulation policies. The combination of both holds the promise to extract crucial contact-related information from the demonstration data and actively exploit it during policy rollouts. However, this integration has so far been underexplored, most notably in dynamic, contact-rich manipulation tasks where precision and reactivity are essential. This work therefore proposes a multimodal, visuotactile imitation learning framework that integrates a modular transformer architecture with a flow-based generative model, enabling efficient learning of fast and dexterous manipulation policies. We evaluate our framework on the dynamic, contact-rich task of robotic match lighting - a task in which tactile feedback influences human manipulation performance. The experimental results highlight the effectiveness of our approach and show that adding tactile information improves policy performance, thereby underlining their combined potential for learning dynamic manipulation from few demonstrations. Project website: https://sites.google.com/view/tactile-il .

Nathan Lambert, Roberto Calandra

Reinforcement learning from human feedback (RLHF) has emerged as a powerful technique to make large language models (LLMs) more capable in complex settings. RLHF proceeds as collecting human preference data, training a reward model on said data, and optimizing a base ML model with respect to said reward for extrinsic evaluation metrics (e.g. MMLU, GSM8k). RLHF relies on many assumptions about how the various pieces fit together, such as a reward model capturing human preferences and an RL optimizer extracting the right signal from a reward model. As the RLHF process involves many distinct design decisions, it is easy to assume that multiple processes are correlated and therefore numerically linked. This apparent correlation is often not true, where reward models are easily overoptimized or RL optimizers can reduce performance on tasks not modeled in the data. Notable manifestations of models trained with imperfect RLHF systems are those that are prone to refusing basic requests for safety reasons or appearing lazy in generations. As chat model evaluation becomes increasingly nuanced, the reliance on a perceived link between reward model training, RL scores, and downstream performance drives these issues, which we describe as an objective mismatch. In this paper, we illustrate the causes of this issue, reviewing relevant literature from model-based reinforcement learning, and argue for solutions. By solving objective mismatch in RLHF, the ML models of the future will be more precisely aligned to user instructions for both safety and helpfulness.

Johannes V. S. Busch, Robert Voelckner, Peter Sossalla et al.

Traffic congestion in dense urban centers presents an economical and environmental burden. In recent years, the availability of vehicle-to-anything communication allows for the transmission of detailed vehicle states to the infrastructure that can be used for intelligent traffic light control. The other way around, the infrastructure can provide vehicles with advice on driving behavior, such as appropriate velocities, which can improve the efficacy of the traffic system. Several research works applied deep reinforcement learning to either traffic light control or vehicle speed advice. In this work, we propose a first attempt to jointly learn the control of both. We show this to improve the efficacy of traffic systems. In our experiments, the joint control approach reduces average vehicle trip delays, w.r.t. controlling only traffic lights, in eight out of eleven benchmark scenarios. Analyzing the qualitative behavior of the vehicle speed advice policy, we observe that this is achieved by smoothing out the velocity profile of vehicles nearby a traffic light. Learning joint control of traffic signaling and speed advice in the real world could help to reduce congestion and mitigate the economical and environmental repercussions of today's traffic systems.

Ran Wei, Nathan Lambert, Anthony McDonald et al.

Model-based Reinforcement Learning (MBRL) aims to make agents more sample-efficient, adaptive, and explainable by learning an explicit model of the environment. While the capabilities of MBRL agents have significantly improved in recent years, how to best learn the model is still an unresolved question. The majority of MBRL algorithms aim at training the model to make accurate predictions about the environment and subsequently using the model to determine the most rewarding actions. However, recent research has shown that model predictive accuracy is often not correlated with action quality, tracing the root cause to the objective mismatch between accurate dynamics model learning and policy optimization of rewards. A number of interrelated solution categories to the objective mismatch problem have emerged as MBRL continues to mature as a research area. In this work, we provide an in-depth survey of these solution categories and propose a taxonomy to foster future research.

Letian Fu, Gaurav Datta, Huang Huang et al.

Touch is an important sensing modality for humans, but it has not yet been incorporated into a multimodal generative language model. This is partially due to the difficulty of obtaining natural language labels for tactile data and the complexity of aligning tactile readings with both visual observations and language descriptions. As a step towards bridging that gap, this work introduces a new dataset of 44K in-the-wild vision-touch pairs, with English language labels annotated by humans (10%) and textual pseudo-labels from GPT-4V (90%). We use this dataset to train a vision-language-aligned tactile encoder for open-vocabulary classification and a touch-vision-language (TVL) model for text generation using the trained encoder. Results suggest that by incorporating touch, the TVL model improves (+29% classification accuracy) touch-vision-language alignment over existing models trained on any pair of those modalities. Although only a small fraction of the dataset is human-labeled, the TVL model demonstrates improved visual-tactile understanding over GPT-4V (+12%) and open-source vision-language models (+32%) on a new touch-vision understanding benchmark. Code and data: https://tactile-vlm.github.io.

Mike Lambeta, Tingfan Wu, Ali Sengul et al.

Touch is a crucial sensing modality that provides rich information about object properties and interactions with the physical environment. Humans and robots both benefit from using touch to perceive and interact with the surrounding environment (Johansson and Flanagan, 2009; Li et al., 2020; Calandra et al., 2017). However, no existing systems provide rich, multi-modal digital touch-sensing capabilities through a hemispherical compliant embodiment. Here, we describe several conceptual and technological innovations to improve the digitization of touch. These advances are embodied in an artificial finger-shaped sensor with advanced sensing capabilities. Significantly, this fingertip contains high-resolution sensors (~8.3 million taxels) that respond to omnidirectional touch, capture multi-modal signals, and use on-device artificial intelligence to process the data in real time. Evaluations show that the artificial fingertip can resolve spatial features as small as 7 um, sense normal and shear forces with a resolution of 1.01 mN and 1.27 mN, respectively, perceive vibrations up to 10 kHz, sense heat, and even sense odor. Furthermore, it embeds an on-device AI neural network accelerator that acts as a peripheral nervous system on a robot and mimics the reflex arc found in humans. These results demonstrate the possibility of digitizing touch with superhuman performance. The implications are profound, and we anticipate potential applications in robotics (industrial, medical, agricultural, and consumer-level), virtual reality and telepresence, prosthetics, and e-commerce. Toward digitizing touch at scale, we open-source a modular platform to facilitate future research on the nature of touch.

Tim Schneider, Guillaume Duret, Cristiana de Farias et al.

Tactile perception has the potential to significantly enhance dexterous robotic manipulation by providing rich local information that can complement or substitute for other sensory modalities such as vision. However, because tactile sensing is inherently local, it is not well-suited for tasks that require broad spatial awareness or global scene understanding on its own. A human-inspired strategy to address this issue is to consider active perception techniques instead. That is, to actively guide sensors toward regions with more informative or significant features and integrate such information over time in order to understand a scene or complete a task. Both active perception and different methods for tactile sensing have received significant attention recently. Yet, despite advancements, both fields lack standardized benchmarks. To bridge this gap, we introduce the Tactile MNIST Benchmark Suite, an open-source, Gymnasium-compatible benchmark specifically designed for active tactile perception tasks, including localization, classification, and volume estimation. Our benchmark suite offers diverse simulation scenarios, from simple toy environments all the way to complex tactile perception tasks using vision-based tactile sensors. Furthermore, we also offer a comprehensive dataset comprising 13,500 synthetic 3D MNIST digit models and 153,600 real-world tactile samples collected from 600 3D printed digits. Using this dataset, we train a CycleGAN for realistic tactile simulation rendering. By providing standardized protocols and reproducible evaluation frameworks, our benchmark suite facilitates systematic progress in the fields of tactile sensing and active perception.

Yves-Simon Zeulner, Sandeep Selvaraj, Roberto Calandra

Towards the grand challenge of achieving human-level manipulation in robots, playing piano is a compelling testbed that requires strategic, precise, and flowing movements. Over the years, several works demonstrated hand-designed controllers on real world piano playing, while other works evaluated robot learning approaches on simulated piano scenarios. In this paper, we develop the first piano playing robotic system that makes use of learning approaches while also being deployed on a real world dexterous robot. Specifically, we make use of Sim2Real to train a policy in simulation using reinforcement learning before deploying the learned policy on a real world dexterous robot. In our experiments, we thoroughly evaluate the interplay between domain randomization and the accuracy of the dynamics model used in simulation. Moreover, we evaluate the robot's performance across multiple songs with varying complexity to study the generalization of our learned policy. By providing a proof-of-concept of learning to play piano in the real world, we want to encourage the community to adopt piano playing as a compelling benchmark towards human-level manipulation. We open-source our code and show additional videos at https://lasr.org/research/learning-to-play-piano .

Tobias Jülg, Pierre Krack, Seongjin Bien et al.

Vision-Language-Action models (VLAs) mark a major shift in robot learning. They replace specialized architectures and task-tailored components of expert policies with large-scale data collection and setup-specific fine-tuning. In this machine learning-focused workflow that is centered around models and scalable training, traditional robotics software frameworks become a bottleneck, while robot simulations offer only limited support for transitioning from and to real-world experiments. In this work, we close this gap by introducing Robot Control Stack (RCS), a lean ecosystem designed from the ground up to support research in robot learning with large-scale generalist policies. At its core, RCS features a modular and easily extensible layered architecture with a unified interface for simulated and physical robots, facilitating sim-to-real transfer. Despite its minimal footprint and dependencies, it offers a complete feature set, enabling both real-world experiments and large-scale training in simulation. Our contribution is twofold: First, we introduce the architecture of RCS and explain its design principles. Second, we evaluate its usability and performance along the development cycle of VLA and RL policies. Our experiments also provide an extensive evaluation of Octo, OpenVLA, and Pi Zero on multiple robots and shed light on how simulation data can improve real-world policy performance. Our code, datasets, weights, and videos are available at: https://robotcontrolstack.github.io/

Mike Lambeta, Huazhe Xu, Jingwei Xu et al.

With the increased availability of rich tactile sensors, there is an equally proportional need for open-source and integrated software capable of efficiently and effectively processing raw touch measurements into high-level signals that can be used for control and decision-making. In this paper, we present PyTouch -- the first machine learning library dedicated to the processing of touch sensing signals. PyTouch, is designed to be modular, easy-to-use and provides state-of-the-art touch processing capabilities as a service with the goal of unifying the tactile sensing community by providing a library for building scalable, proven, and performance-validated modules over which applications and research can be built upon. We evaluate PyTouch on real-world data from several tactile sensors on touch processing tasks such as touch detection, slip and object pose estimations. PyTouch is open-sourced at https://github.com/facebookresearch/pytouch .

Luis Pineda, Brandon Amos, Amy Zhang et al.

Model-based reinforcement learning is a compelling framework for data-efficient learning of agents that interact with the world. This family of algorithms has many subcomponents that need to be carefully selected and tuned. As a result the entry-bar for researchers to approach the field and to deploy it in real-world tasks can be daunting. In this paper, we present MBRL-Lib -- a machine learning library for model-based reinforcement learning in continuous state-action spaces based on PyTorch. MBRL-Lib is designed as a platform for both researchers, to easily develop, debug and compare new algorithms, and non-expert user, to lower the entry-bar of deploying state-of-the-art algorithms. MBRL-Lib is open-source at https://github.com/facebookresearch/mbrl-lib.

Shaoxiong Wang, Mike Lambeta, Po-Wei Chou et al.

Simulators perform an important role in prototyping, debugging, and benchmarking new advances in robotics and learning for control. Although many physics engines exist, some aspects of the real world are harder than others to simulate. One of the aspects that have so far eluded accurate simulation is touch sensing. To address this gap, we present TACTO - a fast, flexible, and open-source simulator for vision-based tactile sensors. This simulator allows to render realistic high-resolution touch readings at hundreds of frames per second, and can be easily configured to simulate different vision-based tactile sensors, including DIGIT and OmniTact. In this paper, we detail the principles that drove the implementation of TACTO and how they are reflected in its architecture. We demonstrate TACTO on a perceptual task, by learning to predict grasp stability using touch from 1 million grasps, and on a marble manipulation control task. Moreover, we provide a proof-of-concept that TACTO can be successfully used for Sim2Real applications. We believe that TACTO is a step towards the widespread adoption of touch sensing in robotic applications, and to enable machine learning practitioners interested in multi-modal learning and control. TACTO is open-source at https://github.com/facebookresearch/tacto.

Mike Lambeta, Po-Wei Chou, Stephen Tian et al.

Despite decades of research, general purpose in-hand manipulation remains one of the unsolved challenges of robotics. One of the contributing factors that limit current robotic manipulation systems is the difficulty of precisely sensing contact forces -- sensing and reasoning about contact forces are crucial to accurately control interactions with the environment. As a step towards enabling better robotic manipulation, we introduce DIGIT, an inexpensive, compact, and high-resolution tactile sensor geared towards in-hand manipulation. DIGIT improves upon past vision-based tactile sensors by miniaturizing the form factor to be mountable on multi-fingered hands, and by providing several design improvements that result in an easier, more repeatable manufacturing process, and enhanced reliability. We demonstrate the capabilities of the DIGIT sensor by training deep neural network model-based controllers to manipulate glass marbles in-hand with a multi-finger robotic hand. To provide the robotic community access to reliable and low-cost tactile sensors, we open-source the DIGIT design at https://digit.ml/.

Sayna Ebrahimi, Franziska Meier, Roberto Calandra et al.

Continual learning aims to learn new tasks without forgetting previously learned ones. We hypothesize that representations learned to solve each task in a sequence have a shared structure while containing some task-specific properties. We show that shared features are significantly less prone to forgetting and propose a novel hybrid continual learning framework that learns a disjoint representation for task-invariant and task-specific features required to solve a sequence of tasks. Our model combines architecture growth to prevent forgetting of task-specific skills and an experience replay approach to preserve shared skills. We demonstrate our hybrid approach is effective in avoiding forgetting and show it is superior to both architecture-based and memory-based approaches on class incrementally learning of a single dataset as well as a sequence of multiple datasets in image classification. Our code is available at \url{https://github.com/facebookresearch/Adversarial-Continual-Learning}.

Elmar Rueckert, Rudolf Lioutikov, Roberto Calandra et al.

Sensor gloves are popular input devices for a large variety of applications including health monitoring, control of music instruments, learning sign language, dexterous computer interfaces, and tele-operating robot hands. Many commercial products as well as low-cost open source projects have been developed. We discuss here how low-cost (approx. 250 EUROs) sensor gloves with force feedback can be build, provide an open source software interface for Matlab and present first results in learning object manipulation skills through imitation learning on the humanoid robot iCub.

Sudharshan Suresh, Haozhi Qi, Tingfan Wu et al.

To achieve human-level dexterity, robots must infer spatial awareness from multimodal sensing to reason over contact interactions. During in-hand manipulation of novel objects, such spatial awareness involves estimating the object's pose and shape. The status quo for in-hand perception primarily employs vision, and restricts to tracking a priori known objects. Moreover, visual occlusion of objects in-hand is imminent during manipulation, preventing current systems to push beyond tasks without occlusion. We combine vision and touch sensing on a multi-fingered hand to estimate an object's pose and shape during in-hand manipulation. Our method, NeuralFeels, encodes object geometry by learning a neural field online and jointly tracks it by optimizing a pose graph problem. We study multimodal in-hand perception in simulation and the real-world, interacting with different objects via a proprioception-driven policy. Our experiments show final reconstruction F-scores of $81$% and average pose drifts of $4.7\,\text{mm}$, further reduced to $2.3\,\text{mm}$ with known CAD models. Additionally, we observe that under heavy visual occlusion we can achieve up to $94$% improvements in tracking compared to vision-only methods. Our results demonstrate that touch, at the very least, refines and, at the very best, disambiguates visual estimates during in-hand manipulation. We release our evaluation dataset of 70 experiments, FeelSight, as a step towards benchmarking in this domain. Our neural representation driven by multimodal sensing can serve as a perception backbone towards advancing robot dexterity. Videos can be found on our project website https://suddhu.github.io/neural-feels/

Haozhi Qi, Brent Yi, Mike Lambeta et al.

Learning policies in simulation and transferring them to the real world has become a promising approach in dexterous manipulation. However, bridging the sim-to-real gap for each new task requires substantial human effort, such as careful reward engineering, hyperparameter tuning, and system identification. In this work, we present a system that leverages low-level skills to address these challenges for more complex tasks. Specifically, we introduce a hierarchical policy for in-hand object reorientation based on previously acquired rotation skills. This hierarchical policy learns to select which low-level skill to execute based on feedback from both the environment and the low-level skill policies themselves. Compared to learning from scratch, the hierarchical policy is more robust to out-of-distribution changes and transfers easily from simulation to real-world environments. Additionally, we propose a generalizable object pose estimator that uses proprioceptive information, low-level skill predictions, and control errors as inputs to estimate the object pose over time. We demonstrate that our system can reorient objects, including symmetrical and textureless ones, to a desired pose.

Tim Schneider, Cristiana de Farias, Roberto Calandra et al.

Active perception is a fundamental skill that enables us humans to deal with uncertainty in our inherently partially observable environment. For senses such as touch, where the information is sparse and local, active perception becomes crucial. In recent years, active perception has emerged as an important research domain in robotics. However, current methods are often bound to specific tasks or make strong assumptions, which limit their generality. To address this gap, this work introduces APPLE (Active Perception Policy Learning) - a novel framework that leverages reinforcement learning (RL) to address a range of different active perception problems. APPLE jointly trains a transformer-based perception module and decision-making policy with a unified optimization objective, learning how to actively gather information. By design, APPLE is not limited to a specific task and can, in principle, be applied to a wide range of active perception problems. We evaluate two variants of APPLE across different tasks, including tactile exploration problems from the Tactile MNIST benchmark. Experiments demonstrate the efficacy of APPLE, achieving high accuracies on both regression and classification tasks. These findings underscore the potential of APPLE as a versatile and general framework for advancing active perception in robotics.

Ken Nakahara, Roberto Calandra

In our daily life, we often encounter objects that are fragile and can be damaged by excessive grasping force, such as fruits. For these objects, it is paramount to grasp gently -- not using the maximum amount of force possible, but rather the minimum amount of force necessary. This paper proposes using visual, tactile, and auditory signals to learn to grasp and regrasp objects stably and gently. Specifically, we use audio signals as an indicator of gentleness during the grasping, and then train an end-to-end action-conditional model from raw visuo-tactile inputs that predicts both the stability and the gentleness of future grasping candidates, thus allowing the selection and execution of the most promising action. Experimental results on a multi-fingered hand over 1,500 grasping trials demonstrated that our model is useful for gentle grasping by validating the predictive performance (3.27% higher accuracy than the vision-only variant) and providing interpretations of their behavior. Finally, real-world experiments confirmed that the grasping performance with the trained multi-modal model outperformed other baselines (17% higher rate for stable and gentle grasps than vision-only). Our approach requires neither tactile sensor calibration nor analytical force modeling, drastically reducing the engineering effort to grasp fragile objects. Dataset and videos are available at https://lasr.org/research/gentle-grasping.

Anas Al Shikh Khalil, Haozhi Qi, Roberto Calandra

As robots become increasingly integrated into everyday tasks, their ability to perceive both the shape and properties of objects during in-hand manipulation becomes critical for adaptive and intelligent behavior. We present SemanticFeels, an extension of the NeuralFeels framework that integrates semantic labeling with neural implicit shape representation, from vision and touch. To illustrate its application, we focus on material classification: high-resolution Digit tactile readings are processed by a fine-tuned EfficientNet-B0 convolutional neural network (CNN) to generate local material predictions, which are then embedded into an augmented signed distance field (SDF) network that jointly predicts geometry and continuous material regions. Experimental results show that the system achieves a high correspondence between predicted and actual materials on both single- and multi-material objects, with an average matching accuracy of 79.87% across multiple manipulation trials on a multi-material object.

Cong Wang, Roberto Calandra, Verena Klös

When robots perform complex and context-dependent tasks in our daily lives, deviations from expectations can confuse users. Explanations of the robot's reasoning process can help users to understand the robot intentions. However, when to provide explanations and what they contain are important to avoid user annoyance. We have investigated user preferences for explanation demand and content for a robot that helps with daily cleaning tasks in a kitchen. Our results show that users want explanations in surprising situations and prefer concise explanations that clearly state the intention behind the confusing action and the contextual factors that were relevant to this decision. Based on these findings, we propose two algorithms to identify surprising actions and to construct effective explanations for Belief-Desire-Intention (BDI) robots. Our algorithms can be easily integrated in the BDI reasoning process and pave the way for better human-robot interaction with context- and user-specific explanations.

Daniel Frau-Alfaro, Julio Castaño-Amoros, Santiago Puente et al.

Object handover is an important skill that we use daily when interacting with other humans. To deploy robots in collaborative setting, like houses, being able to receive and handing over objects safely and efficiently becomes a crucial skill. In this work, we demonstrate the use of Reinforcement Learning (RL) for dexterous object handover between two multi-finger hands. Key to this task is the use of a novel reward function based on dual quaternions to minimize the rotation distance, which outperforms other rotation representations such as Euler and rotation matrices. The robustness of the trained policy is experimentally evaluated by testing w.r.t. objects that are not included in the training distribution, and perturbations during the handover process. The results demonstrate that the trained policy successfully perform this task, achieving a total success rate of 94% in the best-case scenario after 100 experiments, thereby showing the robustness of our policy with novel objects. In addition, the best-case performance of the policy decreases by only 13.8% when the other robot moves during the handover, proving that our policy is also robust to this type of perturbation, which is common in real-world object handovers.

Artemii Redkin, Zdravko Dugonjic, Mike Lambeta et al.

Vision-based tactile sensors use structured light to measure deformation in their elastomeric interface. Until now, vision-based tactile sensors such as DIGIT and GelSight have been using a single, static pattern of structured light tuned to the specific form factor of the sensor. In this work, we investigate the effectiveness of dynamic illumination patterns, in conjunction with image fusion techniques, to improve the quality of sensing of vision-based tactile sensors. Specifically, we propose to capture multiple measurements, each with a different illumination pattern, and then fuse them together to obtain a single, higher-quality measurement. Experimental results demonstrate that this type of dynamic illumination yields significant improvements in image contrast, sharpness, and background difference. This discovery opens the possibility of retroactively improving the sensing quality of existing vision-based tactile sensors with a simple software update, and for new hardware designs capable of fully exploiting dynamic illumination.

Jack Parker-Holder, Raghu Rajan, Xingyou Song et al.

The combination of Reinforcement Learning (RL) with deep learning has led to a series of impressive feats, with many believing (deep) RL provides a path towards generally capable agents. However, the success of RL agents is often highly sensitive to design choices in the training process, which may require tedious and error-prone manual tuning. This makes it challenging to use RL for new problems, while also limits its full potential. In many other areas of machine learning, AutoML has shown it is possible to automate such design choices and has also yielded promising initial results when applied to RL. However, Automated Reinforcement Learning (AutoRL) involves not only standard applications of AutoML but also includes additional challenges unique to RL, that naturally produce a different set of methods. As such, AutoRL has been emerging as an important area of research in RL, providing promise in a variety of applications from RNA design to playing games such as Go. Given the diversity of methods and environments considered in RL, much of the research has been conducted in distinct subfields, ranging from meta-learning to evolution. In this survey we seek to unify the field of AutoRL, we provide a common taxonomy, discuss each area in detail and pose open problems which would be of interest to researchers going forward.

Kevin Sebastian Luck, Roberto Calandra, Michael Mistry

The co-adaptation of robot morphology and behaviour becomes increasingly important with the advent of fast 3D-manufacturing methods and efficient deep reinforcement learning algorithms. A major challenge for the application of co-adaptation methods to the real world is the simulation-to-reality-gap due to model and simulation inaccuracies. However, prior work focuses primarily on the study of evolutionary adaptation of morphologies exploiting analytical models and (differentiable) simulators with large population sizes, neglecting the existence of the simulation-to-reality-gap and the cost of manufacturing cycles in the real world. This paper presents a new approach combining classic high-frequency deep neural networks with computational expensive Graph Neural Networks for the data-efficient co-adaptation of agents with varying numbers of degrees-of-freedom. Evaluations in simulation show that the new method can co-adapt agents within such a limited number of production cycles by efficiently combining design optimization with offline reinforcement learning, that it allows for the direct application to real-world co-adaptation tasks in future work

Edward J. Smith, David Meger, Luis Pineda et al.

Humans build 3D understandings of the world through active object exploration, using jointly their senses of vision and touch. However, in 3D shape reconstruction, most recent progress has relied on static datasets of limited sensory data such as RGB images, depth maps or haptic readings, leaving the active exploration of the shape largely unexplored. Inactive touch sensing for 3D reconstruction, the goal is to actively select the tactile readings that maximize the improvement in shape reconstruction accuracy. However, the development of deep learning-based active touch models is largely limited by the lack of frameworks for shape exploration. In this paper, we focus on this problem and introduce a system composed of: 1) a haptic simulator leveraging high spatial resolution vision-based tactile sensors for active touching of 3D objects; 2)a mesh-based 3D shape reconstruction model that relies on tactile or visuotactile signals; and 3) a set of data-driven solutions with either tactile or visuotactile priors to guide the shape exploration. Our framework enables the development of the first fully data-driven solutions to active touch on top of learned models for object understanding. Our experiments show the benefits of such solutions in the task of 3D shape understanding where our models consistently outperform natural baselines. We provide our framework as a tool to foster future research in this direction.

Huazhe Xu, Yuping Luo, Shaoxiong Wang et al.

The virtuoso plays the piano with passion, poetry and extraordinary technical ability. As Liszt said (a virtuoso)must call up scent and blossom, and breathe the breath of life. The strongest robots that can play a piano are based on a combination of specialized robot hands/piano and hardcoded planning algorithms. In contrast to that, in this paper, we demonstrate how an agent can learn directly from machine-readable music score to play the piano with dexterous hands on a simulated piano using reinforcement learning (RL) from scratch. We demonstrate the RL agents can not only find the correct key position but also deal with various rhythmic, volume and fingering, requirements. We achieve this by using a touch-augmented reward and a novel curriculum of tasks. We conclude by carefully studying the important aspects to enable such learning algorithms and that can potentially shed light on future research in this direction.

Baohe Zhang, Raghu Rajan, Luis Pineda et al.

Model-based Reinforcement Learning (MBRL) is a promising framework for learning control in a data-efficient manner. MBRL algorithms can be fairly complex due to the separate dynamics modeling and the subsequent planning algorithm, and as a result, they often possess tens of hyperparameters and architectural choices. For this reason, MBRL typically requires significant human expertise before it can be applied to new problems and domains. To alleviate this problem, we propose to use automatic hyperparameter optimization (HPO). We demonstrate that this problem can be tackled effectively with automated HPO, which we demonstrate to yield significantly improved performance compared to human experts. In addition, we show that tuning of several MBRL hyperparameters dynamically, i.e. during the training itself, further improves the performance compared to using static hyperparameters which are kept fixed for the whole training. Finally, our experiments provide valuable insights into the effects of several hyperparameters, such as plan horizon or learning rate and their influence on the stability of training and resulting rewards.

Manan Tomar, Amy Zhang, Roberto Calandra et al.

Accuracy and generalization of dynamics models is key to the success of model-based reinforcement learning (MBRL). As the complexity of tasks increases, so does the sample inefficiency of learning accurate dynamics models. However, many complex tasks also exhibit sparsity in the dynamics, i.e., actions have only a local effect on the system dynamics. In this paper, we exploit this property with a causal invariance perspective in the single-task setting, introducing a new type of state abstraction called \textit{model-invariance}. Unlike previous forms of state abstractions, a model-invariance state abstraction leverages causal sparsity over state variables. This allows for compositional generalization to unseen states, something that non-factored forms of state abstractions cannot do. We prove that an optimal policy can be learned over this model-invariance state abstraction and show improved generalization in a simple toy domain. Next, we propose a practical method to approximately learn a model-invariant representation for complex domains and validate our approach by showing improved modelling performance over standard maximum likelihood approaches on challenging tasks, such as the MuJoCo-based Humanoid. Finally, within the MBRL setting we show strong performance gains with respect to sample efficiency across a host of other continuous control tasks.

Nathan O. Lambert, Albert Wilcox, Howard Zhang et al.

Accurately predicting the dynamics of robotic systems is crucial for model-based control and reinforcement learning. The most common way to estimate dynamics is by fitting a one-step ahead prediction model and using it to recursively propagate the predicted state distribution over long horizons. Unfortunately, this approach is known to compound even small prediction errors, making long-term predictions inaccurate. In this paper, we propose a new parametrization to supervised learning on state-action data to stably predict at longer horizons -- that we call a trajectory-based model. This trajectory-based model takes an initial state, a future time index, and control parameters as inputs, and directly predicts the state at the future time index. Experimental results in simulated and real-world robotic tasks show that trajectory-based models yield significantly more accurate long term predictions, improved sample efficiency, and the ability to predict task reward. With these improved prediction properties, we conclude with a demonstration of methods for using the trajectory-based model for control.

Tianyu Li, Roberto Calandra, Deepak Pathak et al.

Hierarchical learning has been successful at learning generalizable locomotion skills on walking robots in a sample-efficient manner. However, the low-dimensional "latent" action used to communicate between two layers of the hierarchy is typically user-designed. In this work, we present a fully-learned hierarchical framework, that is capable of jointly learning the low-level controller and the high-level latent action space. Once this latent space is learned, we plan over continuous latent actions in a model-predictive control fashion, using a learned high-level dynamics model. This framework generalizes to multiple robots, and we present results on a Daisy hexapod simulation, A1 quadruped simulation, and Daisy robot hardware. We compare a range of learned hierarchical approaches from literature, and show that our framework outperforms baselines on multiple tasks and two simulations. In addition to learning approaches, we also compare to inverse-kinematics (IK) acting on desired robot motion, and show that our fully-learned framework outperforms IK in adverse settings on both A1 and Daisy simulations. On hardware, we show the Daisy hexapod achieve multiple locomotion tasks, in an unstructured outdoor setting, with only 2000 hardware samples, reinforcing the robustness and sample-efficiency of our approach.

Luis Pineda, Sumana Basu, Adriana Romero et al.

Deep learning approaches have recently shown great promise in accelerating magnetic resonance image (MRI) acquisition. The majority of existing work have focused on designing better reconstruction models given a pre-determined acquisition trajectory, ignoring the question of trajectory optimization. In this paper, we focus on learning acquisition trajectories given a fixed image reconstruction model. We formulate the problem as a sequential decision process and propose the use of reinforcement learning to solve it. Experiments on a large scale public MRI dataset of knees show that our proposed models significantly outperform the state-of-the-art in active MRI acquisition, over a large range of acceleration factors.

Edward J. Smith, Roberto Calandra, Adriana Romero et al.

When a toddler is presented a new toy, their instinctual behaviour is to pick it upand inspect it with their hand and eyes in tandem, clearly searching over its surface to properly understand what they are playing with. At any instance here, touch provides high fidelity localized information while vision provides complementary global context. However, in 3D shape reconstruction, the complementary fusion of visual and haptic modalities remains largely unexplored. In this paper, we study this problem and present an effective chart-based approach to multi-modal shape understanding which encourages a similar fusion vision and touch information.To do so, we introduce a dataset of simulated touch and vision signals from the interaction between a robotic hand and a large array of 3D objects. Our results show that (1) leveraging both vision and touch signals consistently improves single-modality baselines; (2) our approach outperforms alternative modality fusion methods and strongly benefits from the proposed chart-based structure; (3) there construction quality increases with the number of grasps provided; and (4) the touch information not only enhances the reconstruction at the touch site but also extrapolates to its local neighborhood.

Amy Zhang, Rowan McAllister, Roberto Calandra et al.

We study how representation learning can accelerate reinforcement learning from rich observations, such as images, without relying either on domain knowledge or pixel-reconstruction. Our goal is to learn representations that both provide for effective downstream control and invariance to task-irrelevant details. Bisimulation metrics quantify behavioral similarity between states in continuous MDPs, which we propose using to learn robust latent representations which encode only the task-relevant information from observations. Our method trains encoders such that distances in latent space equal bisimulation distances in state space. We demonstrate the effectiveness of our method at disregarding task-irrelevant information using modified visual MuJoCo tasks, where the background is replaced with moving distractors and natural videos, while achieving SOTA performance. We also test a first-person highway driving task where our method learns invariance to clouds, weather, and time of day. Finally, we provide generalization results drawn from properties of bisimulation metrics, and links to causal inference.

Dieter Büchler, Simon Guist, Roberto Calandra et al.

Dynamic tasks like table tennis are relatively easy to learn for humans but pose significant challenges to robots. Such tasks require accurate control of fast movements and precise timing in the presence of imprecise state estimation of the flying ball and the robot. Reinforcement Learning (RL) has shown promise in learning of complex control tasks from data. However, applying step-based RL to dynamic tasks on real systems is safety-critical as RL requires exploring and failing safely for millions of time steps in high-speed regimes. In this paper, we demonstrate that safe learning of table tennis using model-free Reinforcement Learning can be achieved by using robot arms driven by pneumatic artificial muscles (PAMs). Softness and back-drivability properties of PAMs prevent the system from leaving the safe region of its state space. In this manner, RL empowers the robot to return and smash real balls with 5 m\s and 12m\s on average to a desired landing point. Our setup allows the agent to learn this safety-critical task (i) without safety constraints in the algorithm, (ii) while maximizing the speed of returned balls directly in the reward function (iii) using a stochastic policy that acts directly on the low-level controls of the real system and (iv) trains for thousands of trials (v) from scratch without any prior knowledge. Additionally, we present HYSR, a practical hybrid sim and real training that avoids playing real balls during training by randomly replaying recorded ball trajectories in simulation and applying actions to the real robot. This work is the first to (a) fail-safe learn of a safety-critical dynamic task using anthropomorphic robot arms, (b) learn a precision-demanding problem with a PAM-driven system despite the control challenges and (c) train robots to play table tennis without real balls. Videos and datasets are available at muscularTT.embodied.ml.

Ge Yang, Amy Zhang, Ari S. Morcos et al.

In this paper we introduce plan2vec, an unsupervised representation learning approach that is inspired by reinforcement learning. Plan2vec constructs a weighted graph on an image dataset using near-neighbor distances, and then extrapolates this local metric to a global embedding by distilling path-integral over planned path. When applied to control, plan2vec offers a way to learn goal-conditioned value estimates that are accurate over long horizons that is both compute and sample efficient. We demonstrate the effectiveness of plan2vec on one simulated and two challenging real-world image datasets. Experimental results show that plan2vec successfully amortizes the planning cost, enabling reactive planning that is linear in memory and computation complexity rather than exhaustive over the entire state space.

Suneel Belkhale, Rachel Li, Gregory Kahn et al.

Transporting suspended payloads is challenging for autonomous aerial vehicles because the payload can cause significant and unpredictable changes to the robot's dynamics. These changes can lead to suboptimal flight performance or even catastrophic failure. Although adaptive control and learning-based methods can in principle adapt to changes in these hybrid robot-payload systems, rapid mid-flight adaptation to payloads that have a priori unknown physical properties remains an open problem. We propose a meta-learning approach that "learns how to learn" models of altered dynamics within seconds of post-connection flight data. Our experiments demonstrate that our online adaptation approach outperforms non-adaptive methods on a series of challenging suspended payload transportation tasks. Videos and other supplemental material are available on our website: https://sites.google.com/view/meta-rl-for-flight

Akhil Padmanabha, Frederik Ebert, Stephen Tian et al.

Incorporating touch as a sensing modality for robots can enable finer and more robust manipulation skills. Existing tactile sensors are either flat, have small sensitive fields or only provide low-resolution signals. In this paper, we introduce OmniTact, a multi-directional high-resolution tactile sensor. OmniTact is designed to be used as a fingertip for robotic manipulation with robotic hands, and uses multiple micro-cameras to detect multi-directional deformations of a gel-based skin. This provides a rich signal from which a variety of different contact state variables can be inferred using modern image processing and computer vision methods. We evaluate the capabilities of OmniTact on a challenging robotic control task that requires inserting an electrical connector into an outlet, as well as a state estimation problem that is representative of those typically encountered in dexterous robotic manipulation, where the goal is to infer the angle of contact of a curved finger pressing against an object. Both tasks are performed using only touch sensing and deep convolutional neural networks to process images from the sensor's cameras. We compare with a state-of-the-art tactile sensor that is only sensitive on one side, as well as a state-of-the-art multi-directional tactile sensor, and find that OmniTact's combination of high-resolution and multi-directional sensing is crucial for reliably inserting the electrical connector and allows for higher accuracy in the state estimation task. Videos and supplementary material can be found at https://sites.google.com/berkeley.edu/omnitact

Nathan Lambert, Brandon Amos, Omry Yadan et al.

Model-based reinforcement learning (MBRL) has been shown to be a powerful framework for data-efficiently learning control of continuous tasks. Recent work in MBRL has mostly focused on using more advanced function approximators and planning schemes, with little development of the general framework. In this paper, we identify a fundamental issue of the standard MBRL framework -- what we call the objective mismatch issue. Objective mismatch arises when one objective is optimized in the hope that a second, often uncorrelated, metric will also be optimized. In the context of MBRL, we characterize the objective mismatch between training the forward dynamics model w.r.t.~the likelihood of the one-step ahead prediction, and the overall goal of improving performance on a downstream control task. For example, this issue can emerge with the realization that dynamics models effective for a specific task do not necessarily need to be globally accurate, and vice versa globally accurate models might not be sufficiently accurate locally to obtain good control performance on a specific task. In our experiments, we study this objective mismatch issue and demonstrate that the likelihood of one-step ahead predictions is not always correlated with control performance. This observation highlights a critical limitation in the MBRL framework which will require further research to be fully understood and addressed. We propose an initial method to mitigate the mismatch issue by re-weighting dynamics model training. Building on it, we conclude with a discussion about other potential directions of research for addressing this issue.

Benjamin Letham, Roberto Calandra, Akshara Rai et al.

Bayesian optimization (BO) is a popular approach to optimize expensive-to-evaluate black-box functions. A significant challenge in BO is to scale to high-dimensional parameter spaces while retaining sample efficiency. A solution considered in existing literature is to embed the high-dimensional space in a lower-dimensional manifold, often via a random linear embedding. In this paper, we identify several crucial issues and misconceptions about the use of linear embeddings for BO. We study the properties of linear embeddings from the literature and show that some of the design choices in current approaches adversely impact their performance. We show empirically that properly addressing these issues significantly improves the efficacy of linear embeddings for BO on a range of problems, including learning a gait policy for robot locomotion.

Kevin Sebastian Luck, Heni Ben Amor, Roberto Calandra

Humans and animals are capable of quickly learning new behaviours to solve new tasks. Yet, we often forget that they also rely on a highly specialized morphology that co-adapted with motor control throughout thousands of years. Although compelling, the idea of co-adapting morphology and behaviours in robots is often unfeasible because of the long manufacturing times, and the need to re-design an appropriate controller for each morphology. In this paper, we propose a novel approach to automatically and efficiently co-adapt a robot morphology and its controller. Our approach is based on recent advances in deep reinforcement learning, and specifically the soft actor critic algorithm. Key to our approach is the possibility of leveraging previously tested morphologies and behaviors to estimate the performance of new candidate morphologies. As such, we can make full use of the information available for making more informed decisions, with the ultimate goal of achieving a more data-efficient co-adaptation (i.e., reducing the number of morphologies and behaviors tested). Simulated experiments show that our approach requires drastically less design prototypes to find good morphology-behaviour combinations, making this method particularly suitable for future co-adaptation of robot designs in the real world.

Tianyu Li, Nathan Lambert, Roberto Calandra et al.

Learning to locomote to arbitrary goals on hardware remains a challenging problem for reinforcement learning. In this paper, we present a hierarchical learning framework that improves sample-efficiency and generalizability of locomotion skills on real-world robots. Our approach divides the problem of goal-oriented locomotion into two sub-problems: learning diverse primitives skills, and using model-based planning to sequence these skills. We parametrize our primitives as cyclic movements, improving sample-efficiency of learning on a 18 degrees of freedom robot. Then, we learn coarse dynamics models over primitive cycles and use them in a model predictive control framework. This allows us to learn to walk to arbitrary goals up to 12m away, after about two hours of training from scratch on hardware. Our results on a Daisy hexapod hardware and simulation demonstrate the efficacy of our approach at reaching distant targets, in different environments and with sensory noise.

Thomas Liao, Grant Wang, Brian Yang et al.

Robot design is often a slow and difficult process requiring the iterative construction and testing of prototypes, with the goal of sequentially optimizing the design. For most robots, this process is further complicated by the need, when validating the capabilities of the hardware to solve the desired task, to already have an appropriate controller, which is in turn designed and tuned for the specific hardware. In this paper, we propose a novel approach, HPC-BBO, to efficiently and automatically design hardware configurations, and evaluate them by also automatically tuning the corresponding controller. HPC-BBO is based on a hierarchical Bayesian optimization process which iteratively optimizes morphology configurations (based on the performance of the previous designs during the controller learning process) and subsequently learns the corresponding controllers (exploiting the knowledge collected from optimizing for previous morphologies). Moreover, HPC-BBO can select a "batch" of multiple morphology designs at once, thus parallelizing hardware validation and reducing the number of time-consuming production cycles. We validate HPC-BBO on the design of the morphology and controller for a simulated 6-legged microrobot. Experimental results show that HPC-BBO outperforms multiple competitive baselines, and yields a $360\%$ reduction in production cycles over standard Bayesian optimization, thus reducing the hypothetical manufacturing time of our microrobot from 21 to 4 months.

Dieter Büchler, Roberto Calandra, Jan Peters

High-speed and high-acceleration movements are inherently hard to control. Applying learning to the control of such motions on anthropomorphic robot arms can improve the accuracy of the control but might damage the system. The inherent exploration of learning approaches can lead to instabilities and the robot reaching joint limits at high speeds. Having hardware that enables safe exploration of high-speed and high-acceleration movements is therefore desirable. To address this issue, we propose to use robots actuated by Pneumatic Artificial Muscles (PAMs). In this paper, we present a four degrees of freedom (DoFs) robot arm that reaches high joint angle accelerations of up to 28000 deg/s^2 while avoiding dangerous joint limits thanks to the antagonistic actuation and limits on the air pressure ranges. With this robot arm, we are able to tune control parameters using Bayesian optimization directly on the hardware without additional safety considerations. The achieved tracking performance on a fast trajectory exceeds previous results on comparable PAM-driven robots. We also show that our system can be controlled well on slow trajectories with PID controllers due to careful construction considerations such as minimal bending of cables, lightweight kinematics and minimal contact between PAMs and PAMs with the links. Finally, we propose a novel technique to control the the co-contraction of antagonistic muscle pairs. Experimental results illustrate that choosing the optimal co-contraction level is vital to reach better tracking performance. Through the use of PAM-driven robots and learning, we do a small step towards the future development of robots capable of more human-like motions.

Stephen Tian, Frederik Ebert, Dinesh Jayaraman et al.

Touch sensing is widely acknowledged to be important for dexterous robotic manipulation, but exploiting tactile sensing for continuous, non-prehensile manipulation is challenging. General purpose control techniques that are able to effectively leverage tactile sensing as well as accurate physics models of contacts and forces remain largely elusive, and it is unclear how to even specify a desired behavior in terms of tactile percepts. In this paper, we take a step towards addressing these issues by combining high-resolution tactile sensing with data-driven modeling using deep neural network dynamics models. We propose deep tactile MPC, a framework for learning to perform tactile servoing from raw tactile sensor inputs, without manual supervision. We show that this method enables a robot equipped with a GelSight-style tactile sensor to manipulate a ball, analog stick, and 20-sided die, learning from unsupervised autonomous interaction and then using the learned tactile predictive model to reposition each object to user-specified configurations, indicated by a goal tactile reading. Videos, visualizations and the code are available here: https://sites.google.com/view/deeptactilempc

Justin Lin, Roberto Calandra, Sergey Levine

Much of the literature on robotic perception focuses on the visual modality. Vision provides a global observation of a scene, making it broadly useful. However, in the domain of robotic manipulation, vision alone can sometimes prove inadequate: in the presence of occlusions or poor lighting, visual object identification might be difficult. The sense of touch can provide robots with an alternative mechanism for recognizing objects. In this paper, we study the problem of touch-based instance recognition. We propose a novel framing of the problem as multi-modal recognition: the goal of our system is to recognize, given a visual and tactile observation, whether or not these observations correspond to the same object. To our knowledge, our work is the first to address this type of multi-modal instance recognition problem on such a large-scale with our analysis spanning 98 different objects. We employ a robot equipped with two GelSight touch sensors, one on each finger, and a self-supervised, autonomous data collection procedure to collect a dataset of tactile observations and images. Our experimental results show that it is possible to accurately recognize object instances by touch alone, including instances of novel objects that were never seen during training. Our learned model outperforms other methods on this complex task, including that of human volunteers.

Vicenc Rubies-Royo, Roberto Calandra, Dusan M. Stipanovic et al.

To use neural networks in safety-critical settings it is paramount to provide assurances on their runtime operation. Recent work on ReLU networks has sought to verify whether inputs belonging to a bounded box can ever yield some undesirable output. Input-splitting procedures, a particular type of verification mechanism, do so by recursively partitioning the input set into smaller sets. The efficiency of these methods is largely determined by the number of splits the box must undergo before the property can be verified. In this work, we propose a new technique based on shadow prices that fully exploits the information of the problem yielding a more efficient generation of splits than the state-of-the-art. Results on the Airborne Collision Avoidance System (ACAS) benchmark verification tasks show a considerable reduction in the partitions generated which substantially reduces computation times. These results open the door to improved verification methods for a wide variety of machine learning applications including vision and control.