Heni Ben Amor

Michael Drolet, Simon Stepputtis, Siva Kailas et al.

Amidst the wide popularity of imitation learning algorithms in robotics, their properties regarding hyperparameter sensitivity, ease of training, data efficiency, and performance have not been well-studied in high-precision industry-inspired environments. In this work, we demonstrate the limitations and benefits of prominent imitation learning approaches and analyze their capabilities regarding these properties. We evaluate each algorithm on a complex bimanual manipulation task involving an over-constrained dynamics system in a setting involving multiple contacts between the manipulated object and the environment. While we find that imitation learning is well suited to solve such complex tasks, not all algorithms are equal in terms of handling environmental and hyperparameter perturbations, training requirements, performance, and ease of use. We investigate the empirical influence of these key characteristics by employing a carefully designed experimental procedure and learning environment. Paper website: https://bimanual-imitation.github.io/

Michael Drolet, Joseph Campbell, Heni Ben Amor

We introduce an imitation learning-based physical human-robot interaction algorithm capable of predicting appropriate robot responses in complex interactions involving a superposition of multiple interactions. Our proposed algorithm, Blending Bayesian Interaction Primitives (B-BIP) allows us to achieve responsive interactions in complex hugging scenarios, capable of reciprocating and adapting to a hugs motion and timing. We show that this algorithm is a generalization of prior work, for which the original formulation reduces to the particular case of a single interaction, and evaluate our method through both an extensive user study and empirical experiments. Our algorithm yields significantly better quantitative prediction error and more-favorable participant responses with respect to accuracy, responsiveness, and timing, when compared to existing state-of-the-art methods.

Wen Yang, Zhixian Xie, Yiting Wang et al.

Real-time tracking of previously unseen, highly dynamic objects in contact-rich scenes, such as during dexterous in-hand manipulation, remains a major challenge. Pure vision-based approaches often fail under heavy occlusions due to frequent contact interactions and motion blur caused by abrupt impacts. We propose Twintrack, a physics-aware perception system that enables robust, real-time 6-DoF pose tracking of unknown dynamic objects in contact-rich scenes by leveraging contact physics cues. At its core, Twintrack integrates Real2Sim and Sim2Real. Real2Sim combines vision and contact physics to jointly estimate object geometry and physical properties: an initial reconstruction is obtained from vision, then refined by learning a geometry residual and simultaneously estimating physical parameters (e.g., mass, inertia, and friction) based on contact dynamics consistency. Sim2Real achieves robust pose estimation by adaptively fusing a visual tracker with predictions from the updated contact dynamics. Twintrack is implemented on a GPU-accelerated, customized MJX engine to guarantee real-time performance. We evaluate our method on two contact-rich scenarios: object falling with environmental contacts and multi-fingered in-hand manipulation. Results show that, compared to baselines, Twintrack delivers significantly more robust, accurate, and real-time tracking in these challenging settings, with tracking speeds above 20 Hz. Project page: https://irislab.tech/TwinTrack-webpage/

Sai Krishna Bashetty, Heni Ben Amor, Georgios Fainekos

The goal of this paper is to generate simulations with real-world collision scenarios for training and testing autonomous vehicles. We use numerous dashcam crash videos uploaded on the internet to extract valuable collision data and recreate the crash scenarios in a simulator. We tackle the problem of extracting 3D vehicle trajectories from videos recorded by an unknown and uncalibrated monocular camera source using a modular approach. A working architecture and demonstration videos along with the open-source implementation are provided with the paper.

Lin Guan, Yifan Zhou, Denis Liu et al.

Large-scale generative models are shown to be useful for sampling meaningful candidate solutions, yet they often overlook task constraints and user preferences. Their full power is better harnessed when the models are coupled with external verifiers and the final solutions are derived iteratively or progressively according to the verification feedback. In the context of embodied AI, verification often solely involves assessing whether goal conditions specified in the instructions have been met. Nonetheless, for these agents to be seamlessly integrated into daily life, it is crucial to account for a broader range of constraints and preferences beyond bare task success (e.g., a robot should grasp bread with care to avoid significant deformations). However, given the unbounded scope of robot tasks, it is infeasible to construct scripted verifiers akin to those used for explicit-knowledge tasks like the game of Go and theorem proving. This begs the question: when no sound verifier is available, can we use large vision and language models (VLMs), which are approximately omniscient, as scalable Behavior Critics to catch undesirable robot behaviors in videos? To answer this, we first construct a benchmark that contains diverse cases of goal-reaching yet undesirable robot policies. Then, we comprehensively evaluate VLM critics to gain a deeper understanding of their strengths and failure modes. Based on the evaluation, we provide guidelines on how to effectively utilize VLM critiques and showcase a practical way to integrate the feedback into an iterative process of policy refinement. The dataset and codebase are released at: https://guansuns.github.io/pages/vlm-critic.

Som Sagar, Jiafei Duan, Sreevishakh Vasudevan et al. · uw

Robot manipulation policies often fail for unknown reasons, posing significant challenges for real-world deployment. Researchers and engineers typically address these failures using heuristic approaches, which are not only labor-intensive and costly but also prone to overlooking critical failure modes (FMs). This paper introduces Robot Manipulation Diagnosis (RoboMD), a systematic framework designed to automatically identify FMs arising from unanticipated changes in the environment. Considering the vast space of potential FMs in a pre-trained manipulation policy, we leverage deep reinforcement learning (deep RL) to explore and uncover these FMs using a specially trained vision-language embedding that encodes a notion of failures. This approach enables users to probabilistically quantify and rank failures in previously unseen environmental conditions. Through extensive experiments across various manipulation tasks and algorithms, we demonstrate RoboMD's effectiveness in diagnosing unknown failures in unstructured environments, providing a systematic pathway to improve the robustness of manipulation policies.

Yifan Zhou, Sachin Grover, Mohamed El Mistiri et al.

Reinforcement Learning (RL) traditionally relies on scalar reward signals, limiting its ability to leverage the rich semantic knowledge often available in real-world tasks. In contrast, humans learn efficiently by combining numerical feedback with language, prior knowledge, and common sense. We introduce Prompted Policy Search (ProPS), a novel RL method that unifies numerical and linguistic reasoning within a single framework. Unlike prior work that augment existing RL components with language, ProPS places a large language model (LLM) at the center of the policy optimization loop-directly proposing policy updates based on both reward feedback and natural language input. We show that LLMs can perform numerical optimization in-context, and that incorporating semantic signals, such as goals, domain knowledge, and strategy hints can lead to more informed exploration and sample-efficient learning. ProPS is evaluated across fifteen Gymnasium tasks, spanning classic control, Atari games, and MuJoCo environments, and compared to seven widely-adopted RL algorithms (e.g., PPO, SAC, TRPO). It outperforms all baselines on eight out of fifteen tasks and demonstrates substantial gains when provided with domain knowledge. These results highlight the potential of unifying semantics and numerics for transparent, generalizable, and human-aligned RL.

Keyvan Majd, Siyu Zhou, Heni Ben Amor et al.

In this paper, we propose a framework to repair a pre-trained feed-forward neural network (NN) to satisfy a set of properties. We formulate the properties as a set of predicates that impose constraints on the output of NN over the target input domain. We define the NN repair problem as a Mixed Integer Quadratic Program (MIQP) to adjust the weights of a single layer subject to the given predicates while minimizing the original loss function over the original training domain. We demonstrate the application of our framework in bounding an affine transformation, correcting an erroneous NN in classification, and bounding the inputs of a NN controller.

Shubham Sonawani, Kailas Maneparambil, Heni Ben Amor

Robotic systems for retail have gained a lot of attention due to the labor-intensive nature of such business environments. Many tasks have the potential to be automated via intelligent robotic systems that have manipulation capabilities. For example, empty shelves can be replenished, stray products can be picked up or new items can be delivered. However, many challenges make the realization of this vision a challenge. In particular, robots are still too expensive and do not work out of the box. In this paper, we discuss a work-in-progress approach for enabling power-on-and-go robots in retail environments through a combination of active, physical sensors and passive, artificial sensors. In particular, we use low-cost hardware sensors in conjunction with machine learning techniques in order to generate high-quality environmental information. More specifically, we present a setup in which a standard monocular camera and Bluetooth low-energy yield a reliable robot system that can immediately be used after placing a couple of sensors in the environment. The camera information is used to synthesize accurate 3D point clouds, whereas the BLE data is used to integrate the data into a complex map of the environment. The combination of active and passive sensing enables high-quality sensing capabilities at a fraction of the costs traditionally associated with such tasks.

Geoffrey Clark, Joseph Campbell, Heni Ben Amor

We present Model-Predictive Interaction Primitives -- a robot learning framework for assistive motion in human-machine collaboration tasks which explicitly accounts for biomechanical impact on the human musculoskeletal system. First, we extend Interaction Primitives to enable predictive biomechanics: the prediction of future biomechanical states of a human partner conditioned on current observations and intended robot control signals. In turn, we leverage this capability within a model-predictive control strategy to identify the future ergonomic and biomechanical ramifications of potential robot actions. Optimal control trajectories are selected so as to minimize future physical impact on the human musculoskeletal system. We empirically demonstrate that our approach minimizes knee or muscle forces via generated control actions selected according to biomechanical cost functions. Experiments are performed in synthetic and real-world experiments involving powered prosthetic devices.

Simon Stepputtis, Joseph Campbell, Mariano Phielipp et al.

Imitation learning is a popular approach for teaching motor skills to robots. However, most approaches focus on extracting policy parameters from execution traces alone (i.e., motion trajectories and perceptual data). No adequate communication channel exists between the human expert and the robot to describe critical aspects of the task, such as the properties of the target object or the intended shape of the motion. Motivated by insights into the human teaching process, we introduce a method for incorporating unstructured natural language into imitation learning. At training time, the expert can provide demonstrations along with verbal descriptions in order to describe the underlying intent (e.g., "go to the large green bowl"). The training process then interrelates these two modalities to encode the correlations between language, perception, and motion. The resulting language-conditioned visuomotor policies can be conditioned at runtime on new human commands and instructions, which allows for more fine-grained control over the trained policies while also reducing situational ambiguity. We demonstrate in a set of simulation experiments how our approach can learn language-conditioned manipulation policies for a seven-degree-of-freedom robot arm and compare the results to a variety of alternative methods.

Geoffrey Clark, Joseph Campbell, Seyed Mostafa Rezayat Sorkhabadi et al.

We propose in this paper Periodic Interaction Primitives - a probabilistic framework that can be used to learn compact models of periodic behavior. Our approach extends existing formulations of Interaction Primitives to periodic movement regimes, i.e., walking. We show that this model is particularly well-suited for learning data-driven, customized models of human walking, which can then be used for generating predictions over future states or for inferring latent, biomechanical variables. We also demonstrate how the same framework can be used to learn controllers for a robotic prosthesis using an imitation learning approach. Results in experiments with human participants indicate that Periodic Interaction Primitives efficiently generate predictions and ankle angle control signals for a robotic prosthetic ankle, with MAE of 2.21 degrees in 0.0008s per inference. Performance degrades gracefully in the presence of noise or sensor fall outs. Compared to alternatives, this algorithm functions 20 times faster and performed 4.5 times more accurately on test subjects.

Shubham Sonawani, Ryan Alimo, Renaud Detry et al.

Accurate real-time pose estimation of spacecraft or object in space is a key capability necessary for on-orbit spacecraft servicing and assembly tasks. Pose estimation of objects in space is more challenging than for objects on Earth due to space images containing widely varying illumination conditions, high contrast, and poor resolution in addition to power and mass constraints. In this paper, a convolutional neural network is leveraged to uniquely determine the translation and rotation of an object of interest relative to the camera. The main idea of using CNN model is to assist object tracker used in on space assembly tasks where only feature based method is always not sufficient. The simulation framework designed for assembly task is used to generate dataset for training the modified CNN models and, then results of different models are compared with measure of how accurately models are predicting the pose. Unlike many current approaches for spacecraft or object in space pose estimation, the model does not rely on hand-crafted object-specific features which makes this model more robust and easier to apply to other types of spacecraft. It is shown that the model performs comparable to the current feature-selection methods and can therefore be used in conjunction with them to provide more reliable estimates.

Bijan Fakhri, Troy McDaniel, Heni Ben Amor et al.

As digital worlds become ubiquitous via video games, simulations, virtual and augmented reality, people with disabilities who cannot access those worlds are becoming increasingly disenfranchised. More often than not the design of these environments focuses on vision, making them inaccessible in whole or in part to people with visual impairments. Accessible games and visual aids have been developed but their lack of prevalence or unintuitive interfaces make them impractical for daily use. To address this gap, we present Foveated Haptic Gaze, a method for conveying visual information via haptics that is intuitive and designed for interacting with real-time 3-dimensional environments. To validate our approach we developed a prototype of the system along with a simplified first-person shooter game. Lastly we present encouraging user study results of both sighted and blind participants using our system to play the game with no visual feedback.

Siyu Zhou, Mariano Phielipp, Jorge A. Sefair et al.

In this paper, we propose SwarmNet -- a neural network architecture that can learn to predict and imitate the behavior of an observed swarm of agents in a centralized manner. Tested on artificially generated swarm motion data, the network achieves high levels of prediction accuracy and imitation authenticity. We compare our model to previous approaches for modelling interaction systems and show how modifying components of other models gradually approaches the performance of ours. Finally, we also discuss an extension of SwarmNet that can deal with nondeterministic, noisy, and uncertain environments, as often found in robotics applications.

Simon Stepputtis, Joseph Campbell, Mariano Phielipp et al.

In this work we propose a novel end-to-end imitation learning approach which combines natural language, vision, and motion information to produce an abstract representation of a task, which in turn is used to synthesize specific motion controllers at run-time. This multimodal approach enables generalization to a wide variety of environmental conditions and allows an end-user to direct a robot policy through verbal communication. We empirically validate our approach with an extensive set of simulations and show that it achieves a high task success rate over a variety of conditions while remaining amenable to probabilistic interpretability.

Kevin Sebastian Luck, Mel Vecerik, Simon Stepputtis et al.

Model-free reinforcement learning algorithms such as Deep Deterministic Policy Gradient (DDPG) often require additional exploration strategies, especially if the actor is of deterministic nature. This work evaluates the use of model-based trajectory optimization methods used for exploration in Deep Deterministic Policy Gradient when trained on a latent image embedding. In addition, an extension of DDPG is derived using a value function as critic, making use of a learned deep dynamics model to compute the policy gradient. This approach leads to a symbiotic relationship between the deep reinforcement learning algorithm and the latent trajectory optimizer. The trajectory optimizer benefits from the critic learned by the RL algorithm and the latter from the enhanced exploration generated by the planner. The developed methods are evaluated on two continuous control tasks, one in simulation and one in the real world. In particular, a Baxter robot is trained to perform an insertion task, while only receiving sparse rewards and images as observations from the environment.

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.

Kunal Bagewadi, Joseph Campbell, Heni Ben Amor

A hug is a tight embrace and an expression of warmth, sympathy and camaraderie. Despite the fact that a hug often only takes a few seconds, it is filled with details and nuances and is a highly complex process of coordination between two agents. For human-robot collaborative tasks, it is necessary for humans to develop trust and see the robot as a partner to perform a given task together. Datasets representing agent-agent interaction are scarce and, if available, of limited quality. To study the underlying phenomena and variations in a hug between a person and a robot, we deployed Baxter humanoid robot and wearable sensors on persons to record 353 episodes of hugging activity. 33 people were given minimal instructions to hug the humanoid robot for as natural hugging interaction as possible. In the paper, we present our methodology and analysis of the collected dataset. The use of this dataset is to implement machine learning methods for the humanoid robot to learn to anticipate and react to the movements of a person approaching for a hug. In this regard, we show the significance of the dataset by highlighting certain features in our dataset.

Joseph Campbell, Arne Hitzmann, Simon Stepputtis et al.

Musculoskeletal robots that are based on pneumatic actuation have a variety of properties, such as compliance and back-drivability, that render them particularly appealing for human-robot collaboration. However, programming interactive and responsive behaviors for such systems is extremely challenging due to the nonlinearity and uncertainty inherent to their control. In this paper, we propose an approach for learning Bayesian Interaction Primitives for musculoskeletal robots given a limited set of example demonstrations. We show that this approach is capable of real-time state estimation and response generation for interaction with a robot for which no analytical model exists. Human-robot interaction experiments on a 'handshake' task show that the approach generalizes to new positions, interaction partners, and movement velocities.

Joseph Campbell, Simon Stepputtis, Heni Ben Amor

Human-robot interaction benefits greatly from multimodal sensor inputs as they enable increased robustness and generalization accuracy. Despite this observation, few HRI methods are capable of efficiently performing inference for multimodal systems. In this work, we introduce a reformulation of Interaction Primitives which allows for learning from demonstration of interaction tasks, while also gracefully handling nonlinearities inherent to multimodal inference in such scenarios. We also empirically show that our method results in more accurate, more robust, and faster inference than standard Interaction Primitives and other common methods in challenging HRI scenarios.

Trevor Barron, Oliver Obst, Heni Ben Amor

All reinforcement learning algorithms must handle the trade-off between exploration and exploitation. Many state-of-the-art deep reinforcement learning methods use noise in the action selection, such as Gaussian noise in policy gradient methods or $ε$-greedy in Q-learning. While these methods are appealing due to their simplicity, they do not explore the state space in a methodical manner. We present an approach that uses a model to derive reward bonuses as a means of intrinsic motivation to improve model-free reinforcement learning. A key insight of our approach is that this dynamics model can be learned in the latent feature space of a value function, representing the dynamics of the agent and the environment. This method is both theoretically grounded and computationally advantageous, permitting the efficient use of Bayesian information-theoretic methods in high-dimensional state spaces. We evaluate our method on several continuous control tasks, focusing on improving exploration.

Mark Strickland, Georgios Fainekos, Heni Ben Amor

In this paper, we investigate a predictive approach for collision risk assessment in autonomous and assisted driving. A deep predictive model is trained to anticipate imminent accidents from traditional video streams. In particular, the model learns to identify cues in RGB images that are predictive of hazardous upcoming situations. In contrast to previous work, our approach incorporates (a) temporal information during decision making, (b) multi-modal information about the environment, as well as the proprioceptive state and steering actions of the controlled vehicle, and (c) information about the uncertainty inherent to the task. To this end, we discuss Deep Predictive Models and present an implementation using a Bayesian Convolutional LSTM. Experiments in a simple simulation environment show that the approach can learn to predict impending accidents with reasonable accuracy, especially when multiple cameras are used as input sources.

Kevin Sebastian Luck, Joseph Campbell, Michael Andrew Jansen et al.

We present a methodology for fast prototyping of morphologies and controllers for robot locomotion. Going beyond simulation-based approaches, we argue that the form and function of a robot, as well as their interplay with real-world environmental conditions are critical. Hence, fast design and learning cycles are necessary to adapt robot shape and behavior to their environment. To this end, we present a combination of laminate robot manufacturing and sample-efficient reinforcement learning. We leverage this methodology to conduct an extensive robot learning experiment. Inspired by locomotion in sea turtles, we design a low-cost crawling robot with variable, interchangeable fins. Learning is performed using both bio-inspired and original fin designs in an artificial indoor environment as well as a natural environment in the Arizona desert. The findings of this study show that static policies developed in the laboratory do not translate to effective locomotion strategies in natural environments. In contrast to that, sample-efficient reinforcement learning can help to rapidly accommodate changes in the environment or the robot.

Ana Huaman Quispe, Heni Ben Amor, Henrik Christensen et al.

In this paper we propose an approach for efficient grasp selection for manipulation tasks of unknown objects. Even for simple tasks such as pick-and-place, a unique solution is rare to occur. Rather, multiple candidate grasps must be considered and (potentially) tested till a successful, kinematically feasible path is found. To make this process efficient, the grasps should be ordered such that those more likely to succeed are tested first. We propose to use grasp manipulability as a metric to prioritize grasps. We present results of simulation experiments which demonstrate the usefulness of our metric. Additionally, we present experiments with our physical robot performing simple manipulation tasks with a small set of different household objects.

Samarth Brahmbhatt, Heni Ben Amor, Henrik Christensen

We present a learning approach for localization and segmentation of objects in an image in a manner that is robust to partial occlusion. Our algorithm produces a bounding box around the full extent of the object and labels pixels in the interior that belong to the object. Like existing segmentation aware detection approaches, we learn an appearance model of the object and consider regions that do not fit this model as potential occlusions. However, in addition to the established use of pairwise potentials for encouraging local consistency, we use higher order potentials which capture information at the level of im- age segments. We also propose an efficient loss function that targets both localization and segmentation performance. Our algorithm achieves 13.52% segmentation error and 0.81 area under the false-positive per image vs. recall curve on average over the challenging CMU Kitchen Occlusion Dataset. This is a 42.44% decrease in segmentation error and a 16.13% increase in localization performance compared to the state-of-the-art. Finally, we show that the visibility labelling produced by our algorithm can make full 3D pose estimation from a single image robust to occlusion.