Matei Ciocarlie

Surface electromyography (sEMG) is a promising control signal for assist-as-needed hand rehabilitation after stroke, but detecting intent from paretic muscles often requires lengthy, subject-specific calibration and remains brittle to variability. We propose a healthy-to-stroke adaptation pipeline that initializes an intent detector from a model pretrained on large-scale able-bodied sEMG, then fine-tunes it for each stroke participant using only a small amount of subject-specific data. Using a newly collected dataset from three individuals with chronic stroke, we compare adaptation strategies (head-only tuning, parameter-efficient LoRA adapters, and full end-to-end fine-tuning) and evaluate on held-out test sets that include realistic distribution shifts such as within-session drift, posture changes, and armband repositioning. Across conditions, healthy-pretrained adaptation consistently improves stroke intent detection relative to both zero-shot transfer and stroke-only training under the same data budget; the best adaptation methods improve average transition accuracy from 0.42 to 0.61 and raw accuracy from 0.69 to 0.78. These results suggest that transferring a reusable healthy-domain EMG representation can reduce calibration burden while improving robustness for real-time post-stroke intent detection. Our project website, video, code, and dataset are available at: https://roamlab.github.io/reactemg-stroke/.

Surface electromyography (sEMG) is a promising control signal for assist-as-needed hand rehabilitation after stroke, but detecting intent from paretic muscles often requires lengthy, subject-specific calibration and remains brittle to variability. We propose a healthy-to-stroke adaptation pipeline that initializes an intent detector from a model pretrained on large-scale able-bodied sEMG, then fine-tunes it for each stroke participant using only a small amount of subject-specific data. Using a newly collected dataset from three individuals with chronic stroke, we compare adaptation strategies (head-only tuning, parameter-efficient LoRA adapters, and full end-to-end fine-tuning) and evaluate on held-out test sets that include realistic distribution shifts such as within-session drift, posture changes, and armband repositioning. Across conditions, healthy-pretrained adaptation consistently improves stroke intent detection relative to both zero-shot transfer and stroke-only training under the same data budget; the best adaptation methods improve average transition accuracy from 0.42 to 0.61 and raw accuracy from 0.69 to 0.78. These results suggest that transferring a reusable healthy-domain EMG representation can reduce calibration burden while improving robustness for real-time post-stroke intent detection. Our project website, video, code, and dataset are available at: https://roamlab.github.io/reactemg-stroke/.

In this work, we introduce SpikeATac, a multimodal tactile finger combining a taxelized and highly sensitive dynamic response (PVDF) with a static transduction method (capacitive) for multimodal touch sensing. Named for its `spiky' response, SpikeATac's 16-taxel PVDF film sampled at 4 kHz provides fast, sensitive dynamic signals to the very onset and breaking of contact. We characterize the sensitivity of the different modalities, and show that SpikeATac provides the ability to stop quickly and delicately when grasping fragile, deformable objects. Beyond parallel grasping, we show that SpikeATac can be used in a learning-based framework to achieve new capabilities on a dexterous multifingered robot hand. We use a learning recipe that combines reinforcement learning from human feedback with tactile-based rewards to fine-tune the behavior of a policy to modulate force. Our hardware platform and learning pipeline together enable a difficult dexterous and contact-rich task that has not previously been achieved: in-hand manipulation of fragile objects. Videos are available at https://roamlab.github.io/spikeatac/ .

We introduce GEOTACT, the first robotic system capable of grasping and retrieving objects of potentially unknown shapes buried in a granular environment. While important in many applications, ranging from mining and exploration to search and rescue, this type of interaction with granular media is difficult due to the uncertainty stemming from visual occlusion and noisy contact signals. To address these challenges, we use a learning method relying exclusively on touch feedback, trained end-to-end with simulated sensor noise. We show that our problem formulation leads to the natural emergence of learned pushing behaviors that the manipulator uses to reduce uncertainty and funnel the object to a stable grasp despite spurious and noisy tactile readings. We introduce a training curriculum that bootstraps learning in simulated granular environments, enabling zero-shot transfer to real hardware. Despite being trained only on seven objects with primitive shapes, our method is shown to successfully retrieve 35 different objects, including rigid, deformable, and articulated objects with complex shapes.

The acoustic response of an object can reveal a lot about its global state, for example its material properties or the extrinsic contacts it is making with the world. In this work, we build an active acoustic sensing gripper equipped with two piezoelectric fingers: one for generating signals, the other for receiving them. By sending an acoustic vibration from one finger to the other through an object, we gain insight into an object’s acoustic properties and contact state. We use this system to classify objects, estimate grasping position, estimate poses of internal structures, and classify the types of extrinsic contacts an object is making with the environment. Using our contact type classification model, we tackle a standard long-horizon manipulation problem: peg insertion. We use a simple simulated transition model based on the performance of our sensor to train an imitation learning policy that is robust to imperfect predictions from the classifier. We finally demonstrate the policy on a UR5 robot with active acoustic sensing as the only feedback. Videos can be found at https://roamlab.github.io/vibecheck.

In this paper, we introduce a sensor designed for robotic fingers which can provide information on the displacements induced by external forces. Our sensor uses LEDs to sense the displacement between two plates connected by a transparent elastomer; when a force is applied to the finger, the elastomer displaces and the LED signals change. We show that using LEDs as both light emitters and receivers in this context provides high sensitivity, allowing such an emitter and receiver pairs to detect very small displacements. We characterize the standalone performance of the sensor by testing the ability of a supervised learning model to predict complete force and torque data from its raw signals, and obtain a mean error between 0.05 and 0.07 N across the three directions of force applied to the finger. Our method allows for compact packaging (fitting at the base of a finger) with no amplification electronics, low cost manufacturing, easy integration into a complete hand, and high overload shear forces and bending torques, suggesting future applicability to complete manipulation tasks.

Bimanual robot manipulators can achieve impressive dexterity, but typically rely on two full six- or seven- degree-of-freedom arms so that paired grippers can coordinate effectively. This traditional framework increases system complexity while only exploiting a fraction of the overall workspace for dexterous interaction. We introduce the MiniBEE (Miniature Bimanual End-effector), a compact system in which two reduced-mobility arms (3+ DOF each) are coupled into a kinematic chain that preserves full relative positioning between grippers. To guide our design, we formulate a kinematic dexterity metric that enlarges the dexterous workspace while keeping the mechanism lightweight and wearable. The resulting system supports two complementary modes: (i) wearable kinesthetic data collection with self-tracked gripper poses, and (ii) deployment on a standard robot arm, extending dexterity across its entire workspace. We present kinematic analysis and design optimization methods for maximizing dexterous range, and demonstrate an end-to-end pipeline in which wearable demonstrations train imitation learning policies that perform robust, real-world bimanual manipulation.

Surface electromyography (sEMG) signals show promise for effective human-machine interfaces, particularly in rehabilitation and prosthetics. However, challenges remain in developing systems that respond quickly to user intent, produce stable flicker-free output suitable for device control, and work across different subjects without time-consuming calibration. In this work, we propose a framework for EMG-based intent detection that addresses these challenges. We cast intent detection as per-timestep segmentation of continuous sEMG streams, assigning labels as gestures unfold in real time. We introduce a masked modeling training strategy that aligns muscle activations with their corresponding user intents, enabling rapid onset detection and stable tracking of ongoing gestures. In evaluations against baseline methods, using metrics that capture accuracy, latency and stability for device control, our approach achieves state-of-the-art performance in zero-shot conditions. These results demonstrate its potential for wearable robotics and next-generation prosthetic systems. Our project website, video, code, and dataset are available at: https://reactemg.github.io/

There is growing excitement about the potential of leveraging artificial intelligence (AI) to tackle some of the outstanding barriers to the full deployment of robots in daily lives. However, action and sensing in the physical world pose greater and different challenges for AI than analysing data in isolation and it is important to reflect on which AI approaches are most likely to be successfully applied to robots. Questions to address, among others, are how AI models can be adapted to specific robot designs, tasks and environments. This Perspective offers an assessment of what AI has achieved for robotics since the 1990s and proposes a research roadmap with challenges and promises. These range from keeping up-to-date large datasets, representatives of a diversity of tasks that robots may have to perform, and of environments they may encounter, to designing AI algorithms tailored specifically to robotics problems but generic enough to apply to a wide range of applications and transfer easily to a variety of robotic platforms. For robots to collaborate effectively with humans, they must predict human behaviour without relying on bias-based profiling. Explainability and transparency in AI-driven robot control are essential for building trust, preventing misuse and attributing responsibility in accidents. We close with describing what are, in our view, primary long-term challenges, namely, designing robots capable of lifelong learning, and guaranteeing safe deployment and usage, as well as sustainable development. AI technologies are advancing rapidly, offering new solutions for autonomous robot operation in complex environments. Aude Billard et al. discuss the need to identify and adapt AI technologies for robotics, proposing a research roadmap to address key challenges and opportunities.

Intent inferral, the process by which a robotic device predicts a user's intent from biosignals, offers an effective and intuitive way to control wearable robots. Classical intent inferral methods treat biosignal inputs as unidirectional ground truths for training machine learning models, where the internal state of the model is not directly observable by the user. In this work, we propose reciprocal learning, a bidirectional paradigm that facilitates human adaptation to an intent inferral classifier. Our paradigm consists of iterative, interwoven stages that alternate between updating machine learning models and guiding human adaptation with the use of augmented visual feedback. We demonstrate this paradigm in the context of controlling a robotic hand orthosis for stroke, where the device predicts open, close, and relax intents from electromyographic (EMG) signals and provides appropriate assistance. We use LED progress-bar displays to communicate to the user the predicted probabilities for open and close intents by the classifier. Our experiments with stroke subjects show reciprocal learning improving performance in a subset of subjects (two out of five) without negatively impacting performance on the others. We hypothesize that, during reciprocal learning, subjects can learn to reproduce more distinguishable muscle activation patterns and generate more separable biosignals.

Wearable robotics has the capacity to assist and rehabilitate hand function in stroke survivors. Many devices that use surface electromyography (sEMG) for control rely on extrinsic hand muscle signals, since sEMG sensors are relatively easy to place on the forearm without interfering with hand function. In this work, we target the intrinsic muscles of the thumb, which are superficial to the skin and thus potentially more accessible via sEMG sensing. Traditional, rigid electrodes can not be placed on the hand without adding bulk and affecting hand functionality. We thus present a novel sensing sleeve that uses textile electrodes to measure sEMG activity of intrinsic thumb muscles. We evaluate the sleeve's performance on detecting thumb movements and muscle activity during both isolated and isometric muscle contractions of the thumb and fingers. This work highlights the potential of textile-based sensors as a low-cost, lightweight, and non-obtrusive alternative to conventional sEMG sensors for wearable robotics.

The acoustic response of an object can reveal a lot about its global state, for example its material properties or the extrinsic contacts it is making with the world. In this work, we build an active acoustic sensing gripper equipped with two piezoelectric fingers: one for generating signals, the other for receiving them. By sending an acoustic vibration from one finger to the other through an object, we gain insight into an object's acoustic properties and contact state. We use this system to classify objects, estimate grasping position, estimate poses of internal structures, and classify the types of extrinsic contacts an object is making with the environment. Using our contact type classification model, we tackle a standard long-horizon manipulation problem: peg insertion. We use a simple simulated transition model based on the performance of our sensor to train an imitation learning policy that is robust to imperfect predictions from the classifier. We finally demonstrate the policy on a UR5 robot with active acoustic sensing as the only feedback. Videos can be found at https://roamlab.github.io/vibecheck .

Human-in-the-loop (HitL) robot deployment has gained significant attention in both academia and industry as a semi-autonomous paradigm that enables human operators to intervene and adjust robot behaviors at deployment time, improving success rates. However, continuous human monitoring and intervention can be highly labor-intensive and impractical when deploying a large number of robots. To address this limitation, we propose a method that allows diffusion policies to actively seek human assistance only when necessary, reducing reliance on constant human oversight. To achieve this, we leverage the generative process of diffusion policies to compute an uncertainty-based metric based on which the autonomous agent can decide to request operator assistance at deployment time, without requiring any operator interaction during training. Additionally, we show that the same method can be used for efficient data collection for fine-tuning diffusion policies in order to improve their autonomous performance. Experimental results from simulated and real-world environments demonstrate that our approach enhances policy performance during deployment for a variety of scenarios.

Intent inferral, the process by which a robotic device predicts a user's intent from biosignals, offers an effective and intuitive way to control wearable robots. Classical intent inferral methods treat biosignal inputs as unidirectional ground truths for training machine learning models, where the internal state of the model is not directly observable by the user. In this work, we propose reciprocal learning, a bidirectional paradigm that facilitates human adaptation to an intent inferral classifier. Our paradigm consists of iterative, interwoven stages that alternate between updating machine learning models and guiding human adaptation with the use of augmented visual feedback. We demonstrate this paradigm in the context of controlling a robotic hand orthosis for stroke, where the device predicts open, close, and relax intents from electromyographic (EMG) signals and provides appropriate assistance. We use LED progress-bar displays to communicate to the user the predicted probabilities for open and close intents by the classifier. Our experiments with stroke subjects show reciprocal learning improving performance in a subset of subjects (two out of five) without negatively impacting performance on the others. We hypothesize that, during reciprocal learning, subjects can learn to reproduce more distinguishable muscle activation patterns and generate more separable biosignals.

Wearable robotics have the capacity to assist stroke survivors in assisting and rehabilitating hand function. Many devices that use surface electromyographic (sEMG) for control rely on extrinsic muscle signals, since sEMG sensors are relatively easy to place on the forearm without interfering with hand activity. In this work, we target the intrinsic muscles of the thumb, which are superficial to the skin and thus potentially more accessible via sEMG sensing. However, traditional, rigid electrodes can not be placed on the hand without adding bulk and affecting hand functionality. We thus present a novel sensing sleeve that uses textile electrodes to measure sEMG activity of intrinsic thumb muscles. We evaluate the sleeve's performance on detecting thumb movements and muscle activity during both isolated and isometric muscle contractions of the thumb and fingers. This work highlights the potential of textile-based sensors as a low-cost, lightweight, and non-obtrusive alternative to conventional sEMG sensors for wearable robotics.

Intent inferral on a hand orthosis for stroke patients is challenging due to the difficulty of data collection. Additionally, EMG signals exhibit significant variations across different conditions, sessions, and subjects, making it hard for classifiers to generalize. Traditional approaches require a large labeled dataset from the new condition, session, or subject to train intent classifiers; however, this data collection process is burdensome and time-consuming. In this paper, we propose ChatEMG, an autoregressive generative model that can generate synthetic EMG signals conditioned on prompts (i.e., a given sequence of EMG signals). ChatEMG enables us to collect only a small dataset from the new condition, session, or subject and expand it with synthetic samples conditioned on prompts from this new context. ChatEMG leverages a vast repository of previous data via generative training while still remaining context-specific via prompting. Our experiments show that these synthetic samples are classifier-agnostic and can improve intent inferral accuracy for different types of classifiers. We demonstrate that our complete approach can be integrated into a single patient session, including the use of the classifier for functional orthosis-assisted tasks. To the best of our knowledge, this is the first time an intent classifier trained partially on synthetic data has been deployed for functional control of an orthosis by a stroke survivor.

Individuals with hand paralysis resulting from C6-C7 spinal cord injuries frequently rely on tenodesis for grasping. However, tenodesis generates limited grasping force and demands constant exertion to maintain a grasp, leading to fatigue and sometimes pain. We introduce the MyHand-SCI, a wearable robot that provides grasping assistance through motorized exotendons. Our user-driven device enables independent, ipsilateral operation via a novel Throttle-based Wrist Angle control method, which allows users to maintain grasps without continued wrist extension. A pilot case study with a person with C6 spinal cord injury shows an improvement in functional grasping and grasping force, as well as a preserved ability to modulate grasping force while using our device, thus improving their ability to manipulate everyday objects. This research is a step towards developing effective and intuitive wearable assistive devices for individuals with spinal cord injury.

We propose MetaEMG, a meta-learning approach for fast adaptation in intent inferral on a robotic hand orthosis for stroke. One key challenge in machine learning for assistive and rehabilitative robotics with disabled-bodied subjects is the difficulty of collecting labeled training data. Muscle tone and spasticity often vary significantly among stroke subjects, and hand function can even change across different use sessions of the device for the same subject. We investigate the use of meta-learning to mitigate the burden of data collection needed to adapt high-capacity neural networks to a new session or subject. Our experiments on real clinical data collected from five stroke subjects show that MetaEMG can improve the intent inferral accuracy with a small session- or subject-specific dataset and very few fine-tuning epochs. To the best of our knowledge, we are the first to formulate intent inferral on stroke subjects as a meta-learning problem and demonstrate fast adaptation to a new session or subject for controlling a robotic hand orthosis with EMG signals.

Underactuated manipulators reduce the number of bulky motors, thereby enabling compact and mechanically robust designs. However, fewer actuators than joints means that the manipulator can only access a specific manifold within the joint space, which is particular to a given hardware configuration and can be low-dimensional and/or discontinuous. Determining an appropriate set of hardware parameters for this class of mechanisms, therefore, is difficult - even for traditional task-based co-optimization methods. In this paper, our goal is to implement a task-based design and policy co-optimization method for underactuated, tendon-driven manipulators. We first formulate a general model for an underactuated, tendon-driven transmission. We then use this model to co-optimize a three-link, two-actuator kinematic chain using reinforcement learning. We demonstrate that our optimized tendon transmission and control policy can be transferred reliably to physical hardware with real-world reaching experiments.

In this paper, we introduce a sensor designed for integration in robot fingers, where it can provide information on the displacements induced by external contact. Our sensor uses LEDs to sense the displacement between two plates connected by a transparent elastomer; when a force is applied to the finger, the elastomer displaces and the LED signals change. We show that using LEDs as both light emitters an receivers in this context provides high sensitivity, allowing such an emitter and receiver pairs to detect very small displacements. We characterize the standalone performance of the sensor by testing the ability of a supervised learning model to predict complete force and torque data from its raw signals, and obtain a mean error between 0.05 and 0.07 N across the three directions of force applied to the finger. Our method allows for finger-size packaging with no amplification electronics, low cost manufacturing, easy integration into a complete hand, and high overload shear forces and bending torques, suggesting future applicability to complete manipulation tasks.

Increased effort during use of the paretic arm and hand can provoke involuntary abnormal synergy patterns and amplify stiffness effects of muscle tone for individuals after stroke, which can add difficulty for user-controlled devices to assist hand movement during functional tasks. We study how volitional effort, exerted in an attempt to open or close the hand, affects resistance to robot-assisted movement at the finger level. We perform experiments with three chronic stroke survivors to measure changes in stiffness when the user is actively exerting effort to activate ipsilateral EMG-controlled robot-assisted hand movements, compared with when the fingers are passively stretched, as well as overall effects from sustained active engagement and use. Our results suggest that active engagement of the upper extremity increases muscle tone in the finger to a much greater degree than through passive-stretch or sustained exertion over time. Potential design implications of this work suggest that developers should anticipate higher levels of finger stiffness when relying on user-driven ipsilateral control methods for assistive or rehabilitative devices for stroke.

We present a method for enabling Reinforcement Learning of motor control policies for complex skills such as dexterous manipulation. We posit that a key difficulty for training such policies is the difficulty of exploring the problem state space, as the accessible and useful regions of this space form a complex structure along manifolds of the original high-dimensional state space. This work presents a method to enable and support exploration with Sampling-based Planning. We use a generally applicable non-holonomic Rapidly-exploring Random Trees algorithm and present multiple methods to use the resulting structure to bootstrap model-free Reinforcement Learning. Our method is effective at learning various challenging dexterous motor control skills of higher difficulty than previously shown. In particular, we achieve dexterous in-hand manipulation of complex objects while simultaneously securing the object without the use of passive support surfaces. These policies also transfer effectively to real robots. A number of example videos can also be found on the project website: sbrl.cs.columbia.edu

PURPOSE: Wearable robotic devices are currently being developed to improve upper limb function for individuals with hemiparesis after stroke. Incorporating the views of clinicians during the development of new technologies can help ensure that end products meet clinical needs and can be adopted for patient care. METHODS: In this cross-sectional mixed-methods study, an anonymous online survey was used to gather clinicians' perceptions of a wearable robotic hand orthosis for post-stroke hemiparesis. Participants were asked about their clinical experience and provided feedback on the prototype device after viewing a video. RESULTS: 154 participants completed the survey. Only 18.8% had previous experience with robotic technology. The majority of participants (64.9%) reported that they would use the device for both rehabilitative and assistive purposes. Participants perceived that the device could be used in supervised clinical settings with all phases of stroke. Participants also indicated a need for insurance coverage and quick setup time. CONCLUSIONS: Engaging clinicians early in the design process can help guide the development of wearable robotic devices. Both rehabilitative and assistive functions are valued by clinicians and should be considered during device development. Future research is needed to understand a broader set of stakeholders' perspectives on utility and design.

Intent inferral on a hand orthosis for stroke patients is challenging due to the difficulty of data collection. Additionally, EMG signals exhibit significant variations across different conditions, sessions, and subjects, making it hard for classifiers to generalize. Traditional approaches require a large labeled dataset from the new condition, session, or subject to train intent classifiers; however, this data collection process is burdensome and time-consuming. In this paper, we propose ChatEMG, an autoregressive generative model that can generate synthetic EMG signals conditioned on prompts (i.e., a given sequence of EMG signals). ChatEMG enables us to collect only a small dataset from the new condition, session, or subject and expand it with synthetic samples conditioned on prompts from this new context. ChatEMG leverages a vast repository of previous data via generative training while still remaining context-specific via prompting. Our experiments show that these synthetic samples are classifier-agnostic and can improve intent inferral accuracy for different types of classifiers. We demonstrate that our complete approach can be integrated into a single patient session, including the use of the classifier for functional orthosis-assisted tasks. To the best of our knowledge, this is the first time an intent classifier trained partially on synthetic data has been deployed for functional control of an orthosis by a stroke survivor. Videos, source code, and additional information can be found at https://jxu.ai/chatemg.

Underactuated manipulators reduce the number of bulky motors, thereby enabling compact and mechanically robust designs. However, fewer actuators than joints means that the manipulator can only access a specific manifold within the joint space, which is particular to a given hardware configuration and can be low-dimensional and/or discontinuous. Determining an appropriate set of hardware parameters for this class of mechanisms, therefore, is difficult - even for traditional task-based co-optimization methods. In this paper, our goal is to implement a task-based design and policy co-optimization method for underactuated, tendon-driven manipulators. We first formulate a general model for an underactuated, tendon-driven transmission. We then use this model to co-optimize a three-link, two-actuator kinematic chain using reinforcement learning. We demonstrate that our optimized tendon transmission and control policy can be transferred reliably to physical hardware with real-world reaching experiments.

In a Human-in-the-Loop paradigm, a robotic agent is able to act mostly autonomously in solving a task, but can request help from an external expert when needed. However, knowing when to request such assistance is critical: too few requests can lead to the robot making mistakes, but too many requests can overload the expert. In this paper, we present a Reinforcement Learning based approach to this problem, where a semi-autonomous agent asks for external assistance when it has low confidence in the eventual success of the task. The confidence level is computed by estimating the variance of the return from the current state. We show that this estimate can be iteratively improved during training using a Bellman-like recursion. On discrete navigation problems with both fully-and partially-observable state information, we show that our method makes effective use of a limited budget of expert calls at run-time, despite having no access to the expert at training time.

In this work, we use MEMS microphones as vibration sensors to simultaneously classify texture and estimate contact position and velocity. Vibration sensors are an important facet of both human and robotic tactile sensing, providing fast detection of contact and onset of slip. Microphones are an attractive option for implementing vibration sensing as they offer a fast response and can be sampled quickly, are affordable, and occupy a very small footprint. Our prototype sensor uses only a sparse array (8-9 mm spacing) of distributed MEMS microphones (<$1, 3.76×2.95×1.10 mm) embedded under an elastomer. We use transformer-based architectures for data analysis, taking advantage of the microphones’ high sampling rate to run our models on time-series data as opposed to individual snapshots. This approach allows us to obtain 77.3% average accuracy on 4-class texture classification (84.2% when excluding the slowest drag velocity), 1.8 mm mean error on contact localization, and 5.6 mm/s mean error on contact velocity. We show that the learned texture and localization models are robust to varying velocity and generalize to unseen velocities. We also report that our sensor provides fast contact detection, an important advantage of fast transducers. This investigation illustrates the capabilities one can achieve with a MEMS microphone array alone, leaving valuable sensor real estate available for integration with complementary tactile sensing modalities.

Recently, reinforcement learning has led to dexterous manipulation skills of increasing complexity. Nonetheless, learning these skills in simulation still exhibits poor sample-efficiency which stems from the fact these skills are learned from scratch without the benefit of any domain expertise. In this work, we aim to improve the sample efficiency of learning dexterous in-hand manipulation skills using controllers available via domain knowledge. To this end, we design simple sub-skill controllers and demonstrate improved sample efficiency using a framework that guides exploration toward relevant state space by following actions from these controllers. We are the first to demonstrate learning hard-to-explore finger-gaiting in-hand manipulation skills without the use of an exploratory reset distribution.

We introduce MORPH, a method for co-optimization of hardware design parameters and control policies in simulation using reinforcement learning. Like most co-optimization methods, MORPH relies on a model of the hardware being optimized, usually simulated based on the laws of physics. However, such a model is often difficult to integrate into an effective optimization routine. To address this, we introduce a proxy hardware model, which is always differentiable and enables efficient co-optimization alongside a long-horizon control policy using RL. MORPH is designed to ensure that the optimized hardware proxy remains as close as possible to its realistic counterpart, while still enabling task completion. We demonstrate our approach on simulated 2D reaching and 3D multi-fingered manipulation tasks.

We propose MetaEMG, a meta-learning approach for fast adaptation in intent inferral on a robotic hand orthosis for stroke. One key challenge in machine learning for assistive and rehabilitative robotics with disabled-bodied subjects is the difficulty of collecting labeled training data. Muscle tone and spasticity often vary significantly among stroke subjects, and hand function can even change across different use sessions of the device for the same subject. We investigate the use of meta-learning to mitigate the burden of data collection needed to adapt high-capacity neural networks to a new session or subject. Our experiments on real clinical data collected from five stroke subjects show that MetaEMG can improve the intent inferral accuracy with a small session- or subject-specific dataset and very few fine-tuning epochs. To the best of our knowledge, we are the first to formulate intent inferral on stroke subjects as a meta-learning problem and demonstrate fast adaptation to a new session or subject for controlling a robotic hand orthosis with EMG signals.

Increased effort during use of the paretic arm and hand can provoke involuntary abnormal synergy patterns and amplify stiffness effects of muscle tone for individuals after stroke, which can add difficulty for user-controlled devices to assist hand movement during functional tasks. We study how volitional effort, exerted in an attempt to open or close the hand, affects resistance to robot-assisted movement at the finger level. We perform experiments with three chronic stroke survivors to measure changes in stiffness when the user is actively exerting effort to activate ipsilateral EMG-controlled robot-assisted hand movements, compared with when the fingers are passively stretched, as well as overall effects from sustained active engagement and use. Our results suggest that active engagement of the upper extremity increases muscle tone in the finger to a much greater degree than through passive-stretch or sustained exertion over time. Potential design implications of this work suggest that developers should anticipate higher levels of finger stiffness when relying on user-driven ipsilateral control methods for assistive or rehabilitative devices for stroke.

Individuals with hand paralysis resulting from C6-C7 spinal cord injuries frequently rely on tenodesis for grasping. However, tenodesis generates limited grasping force and demands constant exertion to maintain a grasp, leading to fatigue and sometimes pain. We introduce the MyHand-SCI, a wearable robot that provides grasping assistance through motorized exotendons. Our user-driven device enables independent, ipsilateral operation via a novel Throttle-based Wrist Angle control method, which allows users to maintain grasps without continued wrist extension. A pilot case study with a person with C6 spinal cord injury shows an improvement in functional grasping and grasping force, as well as a preserved ability to modulate grasping force while using our device, thus improving their ability to manipulate everyday objects. This research is a step towards developing effective and intuitive wearable assistive devices for individuals with spinal cord injury.

We introduce GEOTACT, the first robotic system capable of grasping and retrieving objects of potentially unknown shapes buried in a granular environment. While important in many applications, ranging from mining and exploration to search and rescue, this type of interaction with granular media is difficult due to the uncertainty stemming from visual occlusion and noisy contact signals. To address these challenges, we use a learning method relying exclusively on touch feedback, trained end-to-end with simulated sensor noise. We show that our problem formulation leads to the natural emergence of learned pushing behaviors that the manipulator uses to reduce uncertainty and funnel the object to a stable grasp despite spurious and noisy tactile readings. We introduce a training curriculum that bootstraps learning in simulated granular environments, enabling zero-shot transfer to real hardware. Despite being trained only on seven objects with primitive shapes, our method is shown to successfully retrieve 35 different objects, including rigid, deformable, and articulated objects with complex shapes. Videos and additional information can be found at https://jxu.ai/geotact.

We present a method for enabling Reinforcement Learning of motor control policies for complex skills such as dexterous manipulation. We posit that a key difficulty for training such policies is the difficulty of exploring the problem state space, as the accessible and useful regions of this space form a complex structure along manifolds of the original high-dimensional state space. This work presents a method to enable and support exploration with Sampling-based Planning. We use a generally applicable non-holonomic Rapidly-exploring Random Trees algorithm and present multiple methods to use the resulting structure to bootstrap model-free Reinforcement Learning. Our method is effective at learning various challenging dexterous motor control skills of higher difficulty than previously shown. In particular, we achieve dexterous in-hand manipulation of complex objects while simultaneously securing the object without the use of passive support surfaces. These policies also transfer effectively to real robots. A number of example videos can also be found on the project website: https://sbrl.cs.columbia.edu

Restoration of hand function is one of the highest priorities for SCI populations. In this work, we present a prototype of a robotic assistive orthosis capable of implementing tenodesis user control. The underactuated device provides active grasping assistance while preserving free wrist mobility through the use of Bowden cables. This device enables force modulation during grasping, which was effectively leveraged by a participant with C6 SCI to demonstrate improved grasping abilities using the orthosis, scoring 11 on the Grasp and Release Test using the device compared to 1 without it.

In this work, we use MEMS microphones as vibration sensors to simultaneously classify texture and estimate contact position and velocity. Vibration sensors are an important facet of both human and robotic tactile sensing, providing fast detection of contact and onset of slip. Microphones are an attractive option for implementing vibration sensing as they offer a fast response and can be sampled quickly, are affordable, and occupy a very small footprint. Our prototype sensor uses only a sparse array (8-9 mm spacing) of distributed MEMS microphones (&lt;$1, 3.76 x 2.95 x 1.10 mm) embedded under an elastomer. We use transformer-based architectures for data analysis, taking advantage of the microphones' high sampling rate to run our models on time-series data as opposed to individual snapshots. This approach allows us to obtain 77.3% average accuracy on 4-class texture classification (84.2% when excluding the slowest drag velocity), 1.8 mm mean error on contact localization, and 5.6 mm/s mean error on contact velocity. We show that the learned texture and localization models are robust to varying velocity and generalize to unseen velocities. We also report that our sensor provides fast contact detection, an important advantage of fast transducers. This investigation illustrates the capabilities one can achieve with a MEMS microphone array alone, leaving valuable sensor real estate available for integration with complementary tactile sensing modalities.

We introduce MORPH, a method for co-optimization of hardware design parameters and control policies in simulation using reinforcement learning. Like most co-optimization methods, MORPH relies on a model of the hardware being optimized, usually simulated based on the laws of physics. However, such a model is often difficult to integrate into an effective optimization routine. To address this, we introduce a proxy hardware model, which is always differentiable and enables efficient co-optimization alongside a long-horizon control policy using RL. MORPH is designed to ensure that the optimized hardware proxy remains as close as possible to its realistic counterpart, while still enabling task completion. We demonstrate our approach on simulated 2D reaching and 3D multi-fingered manipulation tasks.

In this paper, we present a novel method for achieving dexterous manipulation of complex objects, while simultaneously securing the object without the use of passive support surfaces.We posit that a key difficulty for training such policies in a Reinforcement Learning framework is the difficulty of exploring the problem state space, as the accessible regions of this space form a complex structure along manifolds of a high-dimensional space.To address this challenge, we use two versions of the non-holonomic Rapidly-Exploring Random Trees algorithm; one version is more general, but requires explicit use of the environment's transition function, while the second version uses manipulation-specific kinematic constraints to attain better sample efficiency.In both cases, we use states found via sampling-based exploration to generate reset distributions that enable training control policies under full dynamic constraints via model-free Reinforcement Learning.We show that these policies are effective at manipulation problems of higher difficulty than previously shown, and also transfer effectively to real robots.

Tactile recognition of 3D objects remains a challenging task. Compared to 2D shapes, the complex geometry of 3D surfaces requires richer tactile signals, more dexterous actions, and more advanced encoding techniques. In this work, we propose TANDEM3D, a method that applies a co-training framework for exploration and decision making to 3D object recognition with tactile signals. Starting with our previous work, which introduced a co-training paradigm for 2D recognition problems, we introduce a number of advances that enable us to scale up to 3D. TANDEM3D is based on a novel encoder that builds 3D object representation from contact positions and normals using PointNet++. Furthermore, by enabling 6DOF movement, TANDEM3D explores and collects discriminative touch information with high efficiency. Our method is trained entirely in simulation and validated with real-world experiments. Compared to state-of-the-art baselines, TANDEM3D achieves higher accuracy and a lower number of actions in recognizing 3D objects and is also shown to be more robust to different types and amounts of sensor noise.

In a Human-in-the-Loop paradigm, a robotic agent is able to act mostly autonomously in solving a task, but can request help from an external expert when needed. However, knowing when to request such assistance is critical: too few requests can lead to the robot making mistakes, but too many requests can overload the expert. In this paper, we present a Reinforcement Learning based approach to this problem, where a semi-autonomous agent asks for external assistance when it has low confidence in the eventual success of the task. The confidence level is computed by estimating the variance of the return from the current state. We show that this estimate can be iteratively improved during training using a Bellman-like recursion. On discrete navigation problems with both fully- and partially-observable state information, we show that our method makes effective use of a limited budget of expert calls at run-time, despite having no access to the expert at training time.