CHI ’25, Yokohama, Japan

This paper introduces ActionSense, a multimodal dataset and recording framework with an emphasis on wearable sensing in a kitchen environment. It provides rich, synchronized data streams along with ground truth data to facilitate learning pipelines that could extract insights about how humans interact with the physical world during activities of daily living, and help lead to more capable and collaborative robot assistants. The wearable sensing suite captures motion, force, and attention information; it includes eye tracking with a first-person camera, forearm muscle activity sensors, a body-tracking system using 17 inertial sensors, finger-tracking gloves, and custom tactile sensors on the hands that use a matrix of conductive threads. This is coupled with activity labels and with externally-captured data from multiple RGB cameras, a depth camera, and microphones. The specific tasks recorded in ActionSense are designed to highlight lower-level physical skills and higher-level scene reasoning or action planning. They include simple object manipulations (e.g., stacking plates), dexterous actions (e.g., peeling or cutting vegetables), and complex action sequences (e.g., setting a table or loading a dishwasher). The resulting dataset and underlying experiment framework are available at https://action-sense.csail.mit.edu. Preliminary networks and analyses explore modality subsets and cross-modal correlations. ActionSense aims to support applications including learning from demonstrations, dexterous robot control, cross-modal predictions, and fine-grained action segmentation. It could also help inform the next generation of smart textiles that may one day unobtrusively send rich data streams to in-home collaborative or autonomous robot assistants.

As the capacity for machines to extend human capabilities continues to grow, the communication channels used must also expand. Allowing machines to interpret nonverbal commands such as gestures can help make interactions more similar to interactions with another person. Yet to be pervasive and effective in realistic scenarios, such interfaces should not require significant sensing infrastructure or per-user setup time. The presented work takes a step towards these goals by using wearable muscle and motion sensors to detect gestures without dedicated calibration or training procedures. An algorithm is presented for clustering unlabeled streaming data in real time, and it is applied to adaptively thresholding muscle and motion signals acquired via electromyography (EMG) and an inertial measurement unit (IMU). This enables plug-and-play online detection of arm stiffening, fist clenching, rotation gestures, and forearm activation. It also augments a neural network pipeline, trained only on strategically chosen training data from previous users, to detect left, right, up, and down gestures. Together, these pipelines offer a plug-and-play gesture vocabulary suitable for remotely controlling a robot. Experiments with 6 subjects evaluate classifier performance and interface efficacy. Classifiers correctly identified 97.6% of 1,200 cued gestures, and a drone correctly responded to 81.6% of 1,535 unstructured gestures as subjects remotely controlled it through target hoops during 119 minutes of total flight time.

As artificial intelligence becomes an increasingly prevalent method of enhancing robotic capabilities, it is important to consider effective ways to train these learning pipelines and to leverage human expertise. Working towards these goals, a master-apprentice model is presented and is evaluated during a grasping task for effectiveness and human perception. The apprenticeship model augments self-supervised learning with learning by demonstration, efficiently using the human's time and expertise while facilitating future scalability to supervision of multiple robots; the human provides demonstrations via virtual reality when the robot cannot complete the task autonomously. Experimental results indicate that the robot learns a grasping task with the apprenticeship model faster than with a solely self-supervised approach and with fewer human interventions than a solely demonstration-based approach; 100% grasping success is obtained after 150 grasps with 19 demonstrations. Preliminary user studies evaluating workload, usability, and effectiveness of the system yield promising results for system scalability and deployability. They also suggest a tendency for users to overestimate the robot's skill and to generalize its capabilities, especially as learning improves.

Seamless communication of desired motions and goals is essential for enabling effective physical human-robot collaboration. In such cases, muscle activity measured via surface electromyography (EMG) can provide insight into a person's intentions while minimally distracting from the task. The presented system uses two muscle signals to create a control framework for team lifting tasks in which a human and robot lift an object together. A continuous setpoint algorithm uses biceps activity to estimate changes in the user's hand height, and also allows the user to explicitly adjust the robot by stiffening or relaxing their arm. In addition to this pipeline, a neural network trained only on previous users classifies biceps and triceps activity to detect up or down gestures on a rolling basis; this enables finer control over the robot and expands the feasible workspace. The resulting system is evaluated by 10 untrained subjects performing a variety of team lifting and assembly tasks with rigid and flexible objects.

Communication with a robot using brain activity from a human collaborator could provide a direct and fast feedback loop that is easy and natural for the human, thereby enabling a wide variety of intuitive interaction tasks. This paper explores the application of EEG-measured error-related potentials (ErrPs) to closed-loop robotic control. ErrP signals are particularly useful for robotics tasks because they are naturally occurring within the brain in response to an unexpected error. We decode ErrP signals from a human operator in real time to control a Rethink Robotics Baxter robot during a binary object selection task. We also show that utilizing a secondary interactive error-related potential signal generated during this closed-loop robot task can greatly improve classification performance, suggesting new ways in which robots can acquire human feedback. The design and implementation of the complete system is described, and results are presented for realtime closed-loop and open-loop experiments as well as offline analysis of both primary and secondary ErrP signals. These experiments are performed using general population subjects that have not been trained or screened. This work thereby demonstrates the potential for EEG-based feedback methods to facilitate seamless robotic control, and moves closer towards the goal of real-time intuitive interaction.

Computational thinking is an important part of a modern education, and robotics provides a powerful tool for teaching programming logic in an interactive and engaging way. The robot garden presented in this paper is a distributed multi-robot system capable of running autonomously or under user control from a simple graphical interface. Over 100 origami flowers are actuated with LEDs and printed pouch motors, and are deployed in a modular array around additional swimming and crawling folded robots. The garden integrates state-of-the-art rapid design and fabrication technologies with distributed systems software techniques to create a scalable swarm in which robots can be controlled individually or as a group. The garden can be used to teach basic algorithmic concepts through its distributed algorithm demonstration capabilities and can teach programming concepts through its education-oriented user interface.

Designing and fabricating new robotic systems is typically limited to experts, requiring engineering background, expensive tools, and considerable time. In contrast, to facilitate everyday users developing custom robots for personal use, this work presents a new system to easily create printable foldable robots from high-level structural specifications. A user merely needs to select electromechanical components from a library of basic building blocks and pre-designed mechanisms, then connect them to define custom robot assemblies. The system then generates complete mechanical drawings suitable for fabrication, instructions for the assembly of electronics, and software to control and drive the final robot. Several robots designed in this manner demonstrate the ability and versatility of this process.

This paper describes a novel underactuated robotic hand design. The hand is highly underactuated as it contains three fingers with three joints each controlled by a single motor. One of the fingers (“thumb”) can also be rotated about the base of the hand, yielding a total of two controllable degrees-of-freedom. A key component of the design is the addition of position and tactile sensors which provide precise angle feedback and binary force feedback. Our mechanical design can be analyzed theoretically to predict contact forces as well as hand position given a particular object shape