Robotics

Rus has contributed some of the first multi-robot system algorithms with performance guarantees in distributed robotics, by introducing a control-theoretic optimization approach for adaptive decentralized coordination.^[8] Key to these results is the tight coupling between perception, control, and communication. The control algorithms are decentralized, adaptive, and provably stable.

Her group has developed self-configuring modular robots that can alter their physical structures to perform different tasks. This includes sets of robotic cubes that use angular movement to assemble into different formations,^[9] and magnet-controlled robots that can walk, sail and glide using different dissolvable exoskeletons.^[10] She has also worked on algorithms for robots to fly in swarms,^[11] and for boats to autonomously navigate the canals of Amsterdam & self-assemble as floating structures.^[12]

Rus was an early contributor to the field of soft robotics, which some researchers believe has the potential to outperform traditional hard-bodied robotics in a range of human environments.^[13] Her work has introduced self-contained autonomous robotic systems such as an underwater “fish” used for ocean exploration ^[14] and dexterous hands that can grasp a range of different objects.^[15] Rus has created inexpensive designs and fabrication techniques for a range of silicon-based robots and 3D-printable robots,^[16] with the goal of making it easier for non-experts to make their own.

Her projects have often drawn inspiration from nature, including the robotic fish and a trunk-like robot imbued with touch sensors.^[17] She has also explored the potential of extremely small-scale robots, like an ingestible origami robot^[18] that could unfold in a person's stomach to patch wounds. Other work has revolved around robots for a range of logistics environments, including one that can disinfect a warehouse floor in 30 minutes.^[19]

Machine learning

Rus and her team are trying to address some of the key challenges with today's methods for machine learning, including data quality and bias, explainability, generalizability, and sustainability. She is working on a new class of machine learning models that she calls “liquid networks” that can more accurately estimate uncertainty,^[20] better understand the cause-and-effect of tasks,^[21] and even that can continuously adapt to new data inputs^[22] rather than only learning during the training phase. Rus' research has also involved developing machine learning systems for a range of use cases and industries, including for autonomous technologies for vehicles on land, in the air and at sea. She has worked on algorithms to improve autonomous driving in difficult road conditions, from country roads^[23] to snowy weather,^[24] and also released an open-source simulation engine that researchers can use to test their algorithms for autonomous vehicles.

Human/robot interaction

Many of the Distributed Robotics Lab's projects have focused on enabling smoother and more natural interaction and collaboration between humans and robots. Rus has created feedback systems that allow human users to subconsciously communicate through brainwave activity whether a robot has made a mistake in manufacturing environments.^[25] Using wearable body sensors, she has developed systems that enable users to more smoothly control drones^[26] and work with to lift and transport goods.^[27]

Her group has also worked on projects geared towards helping the physically disabled. They have collaborated with the Andrea Bocelli Foundation to create wearable systems ^[28] to help guide the visually impaired, as well as a “smart glove” that uses machine learning to interpret sign language.^[29]

Computational design and fabrication

In recent years Rus has worked with MIT colleague Wojciech Matusik to create methods for 3D-printing robots and other functional objects, often made out of multiple different types of material. She has 3D-printed soft robots with embedded electronics,^[30] items with tunable mechanical properties,^[31] and even “smart gloves” that could help with grasping tasks for people with motor-coordination issues.^[32] Her group has developed methods for 3D-printing materials to sense how they are moving and interacting with their environment, which could be used to create soft robots that have some sort of understanding of their own posture and movements.