Daniela Rus

Multi-agent reinforcement learning (MARL) research is faced with a trade-off: it either uses complex environments requiring large compute resources, which makes it inaccessible to researchers with limited resources, or relies on simpler dynamics for faster execution, which makes the transferability of the results to more realistic tasks challenging. Motivated by these challenges, we present Gigastep, a fully vectorizable, MARL environment implemented in JAX, capable of executing up to one billion environment steps per second on consumer-grade hardware. Its design allows for comprehensive MARL experimentation, including a complex, high-dimensional space defined by 3D dynamics, stochasticity, and partial observations. Gigastep supports both collaborative and adversarial tasks, continuous and discrete action spaces, and provides RGB image and feature vector observations, allowing the evaluation of a wide range of MARL algorithms. We validate Gigastep's usability through an extensive set of experiments, underscoring its role in widening participation and promoting inclusivity in the MARL research community.

Integrating Brain-Machine Interfaces into non-clinical applications like robot motion control remains difficult - despite remarkable advancements in clinical settings. Specifically, EEG-based motor imagery systems are still error-prone, posing safety risks when rigid robots operate near humans. This work presents an alternative pathway towards safe and effective operation by combining wearable EEG with physically embodied safety in soft robots. We introduce and test a pipeline that allows a user to move a soft robot's end effector in real time via brain waves that are measured by as few as three EEG channels. A robust motor imagery algorithm interprets the user's intentions to move the position of a virtual attractor to which the end effector is attracted, thanks to a new Cartesian impedance controller. We specifically focus here on planar soft robot-based architected metamaterials, which require the development of a novel control architecture to deal with the peculiar nonlinearities - e.g., non-affinity in control. We preliminarily but quantitatively evaluate the approach on the task of setpoint regulation. We observe that the user reaches the proximity of the setpoint in 66% of steps and that for successful steps, the average response time is 21.5s. We also demonstrate the execution of simple real-world tasks involving interaction with the environment, which would be extremely hard to perform if it were not for the robot's softness.

An autonomous, mobile robotic device (AMR) is configured with one or more UVC radiation sources, and operates to traverse a path while disinfecting an interior space. Each UVC radiation source is connected to the AMR by an articulating arm that is controlled to orient each source towards a feature or surface that is selected for disinfection during the time that the AMR is moving through the space. The location of each feature selected for disinfection can be mapped, and this map information, a current AMR location and pose can be used to generate signals that are used to control the articulating arm to orient each UVC lamp towards a feature that is selected for disinfection.

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

Understanding and anticipating intraoperative events and actions is critical for intraoperative assistance and decision-making during minimally invasive surgery. Automated prediction of events, actions, and the following consequences is addressed through various computational approaches with the objective of augmenting surgeons' perception and decision-making capabilities. We propose a predictive neural network that is capable of understanding and predicting critical interactive aspects of surgical workflow from intra-abdominal video, while flexibly leveraging surgical knowledge graphs. The approach incorporates a hypergraph-transformer (HGT) structure that encodes expert knowledge into the network design and predicts the hidden embedding of the graph. We verify our approach on established surgical datasets and applications, including the detection and prediction of action triplets, and the achievement of the Critical View of Safety (CVS). Moreover, we address specific, safety-related tasks, such as predicting the clipping of cystic duct or artery without prior achievement of the CVS. Our results demonstrate the superiority of our approach compared to unstructured alternatives.

Sensing plays a pivotal role in robotic manipulation, dictating the accuracy and versatility with which objects are handled. Vision-based sensing methods often suffer from fabrication complexity and low durability, while approaches that rely on direct measurements on the gripper often have limited resolution and are difficult to scale. Here, we present a soft robotic gripper made out of two cubic lattices that are sensorized by embedding air channels within the structure. The lattices are 3D printed from a single build material, simplifying the fabrication process. The flexibility of this approach offers significant control over sensor and lattice design, while the pressure-based internal sensing provides measurements with minimal disruption to the grasping surface. With only 12 sensors, 6 per lattice, this gripper can estimate an object’s weight and location and offer new insights into grasp parameters like friction coefficients and grasp force.

In some aspects, a system comprises a computer hardware processor and a non-transitory computer-readable storage medium storing processor-executable instructions for receiving, from one or more sensors, sensor data relating to a robot; generating, using a statistical model, based on the sensor data, first control information for the robot to accomplish a task; transmitting, to the robot, the first control information for execution of the task; and receiving, from the robot, a result of execution of the task.

The sports industry is witnessing an increasing trend of utilizing multiple synchronized sensors for player data collection, enabling personalized training systems with multi-perspective real-time feedback. Badminton could benefit from these various sensors, but there is a scarcity of comprehensive badminton action datasets for analysis and training feedback. Addressing this gap, this paper introduces a multi-sensor badminton dataset for forehand clear and backhand drive strokes, based on interviews with coaches for optimal usability. The dataset covers various skill levels, including beginners, intermediates, and experts, providing resources for understanding biomechanics across skill levels. It encompasses 7,763 badminton swing data from 25 players, featuring sensor data on eye tracking, body tracking, muscle signals, and foot pressure. The dataset also includes video recordings, detailed annotations on stroke …