Researchers work on everything from building better robots to reimagining urban landscapes. Here’s a glimpse at just some of the design work underway.
Elevating the health of our cities
“The resilience of our cities and human landscapes is the biggest design challenge of the 21st century,” says Nicholas de Monchaux, professor and head of architecture at MIT. That’s why de Monchaux is creating a suite of digital tools to help cities weather the challenges presented by climate change, in part by optimizing overlooked urban spaces such as alleys and vacant lots. If such remnant parcels were landscaped, they could aid in drainage and mitigate air pollution, he explains.
“Cities are organic living things, just like a plant or a human body,” he says. “You need to elevate the health of what’s there.”
In an ongoing project begun at the University of California at Berkeley, de Monchaux has digitally mapped thousands of abandoned or underused urban sites and created site-specific designs for park-like community spaces that can mitigate ecological threats from storm flooding, heat waves, and air pollution. Specific design proposals for 3,659 individual sites were published in de Monchaux’s book Local Code (Princeton Architectural Press, 2016).
The aim of this work is to link urban health with social welfare in a “distributed immune system for the 21st century city,” says de Monchaux.
Collaborating with Carlos Sandoval Olascoaga SM ’16, PhD ’21, a former Berkeley colleague who is now a postdoctoral associate and lecturer in MIT’s School of Architecture and Planning, de Monchaux is employing the same digital mapping techniques to identify New York City properties that might be used as urban farms.
In this work, both researchers say they hope to promote the democratization of design by enabling community members to collaborate in selecting farm sites. Their mapping software allows non-technical users to evaluate the overall social and ecological impact of a site by graphically synthesizing data analytics and providing interactive feedback. Sandoval Olascoaga says he hopes the work leads to cities that are shaped “collectively and holistically, rather than from the top down.”
Democratizing product creation with AI
Thomas Edison once said: “I find out what the world needs. Then I go ahead and try to invent it.”
Faez Ahmed seeks to harness the power of artificial intelligence (AI) to help people connect with their inner Edisons.
Ahmed and his team at the Department of Mechanical Engineering’s Design Computation and Digital Engineering (DeCoDE) Lab are creating new AI-driven methods to generate novel designs for various products, from bicycles to aircrafts and ships. Using a database of designs created by people as a starting point, the team applies new machine-learning algorithms to identify promising elements and then uses computer simulations for accelerated discovery.
“The vision of the lab is to create a world where humans and AI design together to solve some of our biggest challenges,” says Ahmed, the Brit and Alex d’Arbeloff Career Development Professor in Engineering Design and assistant professor of mechanical engineering.
When humans design new ships or bicycles, he says, the tweaks they make tend to be incremental. “They take a design and maybe change it a little bit. They may not think out of the box,” Ahmed says. Also, it is difficult for people to consider millions of options.
This is where Ahmed’s method comes in. “We create new algorithms that can learn from humans and then work with humans to create better product designs,” he says.
Humans guide AI to tailor and perfect their designs, he explains. “Right now, designs are created at the headquarters of major industrial companies,” he says. With AI-enabled design democratization, “even if you’re not trained in engineering methods, you can create designs.”
For example, working with a dataset of thousands of bicycle designs made by bicycle enthusiasts, the group devised machine-learning tools to distinguish between styles, functions, and parts. The tools leverage this knowledge to generate innovative new bike designs that meet customer needs. Ahmed’s team is now working to build algorithms that can create designs worthy of being patented.
Above: With vast applications in machine design, a mechanical linkage mechanism—such as this complex version—translates one type of motion into another. Working in assistant professor Faez Ahmed’s DeCoDE lab and in collaboration with the MIT-IBM Watson AI Lab, doctoral student Amin Heyrani Nobari SM ’22 created LINKS, a dataset of a hundred million planar linkage mechanisms aimed at enabling high-performing data-driven models and helping engineers optimize their designs.
Exploring sonic possibilities
Engineering is a creative process, says Ian Hattwick, who helps MIT students tap this creativity to make music in his class 21M.370 Digital Instrument Design. Technology is central to the artistic practice of the class, in which students examine aspects of sound and learn to build instruments through the design of software systems, hardware interfaces, or interactive artworks.
“There are very tangible qualities to sound,” says Hattwick, an engineer, professional musician, and a lecturer in Music and Theater Arts. “It’s fun to explore but unpredictable. Students get a handle on that unpredictability and interact with emerging sonic processes and sonic results.” Successful instrument building, he points out, requires interdisciplinary skills, including multimedia software programming and electronic and mechanical engineering.
During the 2020–2021 school year, Hattwick invited professional musicians who incorporate digital and electronic musical instruments into their practice to participate in an online concert, “Engineered Expressions.” Students enjoyed performances by Myriam Bleau, Marije Baalman, 80KV, and Author & Punisher, then had the opportunity to engage with the artists in a virtual workshop.
Digital Instrument Design students study the work of such contemporary musicians, then design, build, and play instruments consisting of electronic, mechanical, and software components. Many work on their designs in the Voxel Lab, the music and art makerspace in the Institute’s new InnovationHQ in Kendall Square. The lab is also headquarters for FaMLE, the MIT Laptop Ensemble, which under Hattwick’s direction explores emerging digital music practices such as live coding, a practice in which performers write code in real time to generate music and visuals.
Hattwick enjoys watching students gain artistic and technical confidence as the Digital Instrument Design course progresses. “They should feel empowered to make decisions, think through the implications of the decisions they’re making, and follow their interests in the things they create.”
Creating robots with human-level perception
“This is a great time for robotics,” says Luca Carlone, the Leonardo Career Development Professor in Engineering and associate professor of aeronautics and astronautics. “What we do can have a real impact on the world.”
Carlone and his team at the SPARK (Sensing, Perception, Autonomy, and Robot Kinetics) Lab are working to help robots navigate the world with ease, the way humans do. “As humans, we form a complex internal model of the external world, which we use to navigate and make decisions,” says Carlone. “Similarly, robots need spatial perception algorithms to understand their surroundings.”
A lack of scene understanding by an autonomous vehicle can lead to failures—for example, a self-driving car may crash as a result of incorrectly interpreting its sensor data. SPARK Lab researchers are developing robust and certifiable algorithms and systems that enable robots to understand 3-D scenes. They have also developed groundbreaking tools to build hierarchical models (called 3-D scene graphs) of the environment as the robot navigates in it. “My lab is about pushing the state-of-the-art in perception, increasing the robustness of the algorithms, and getting performance guarantees,” says Carlone.
SPARK Lab is also working on the flight and grasping capabilities of robots and drones, since many aerial manipulators are still relatively clumsy. Carlone’s goal is to take lessons from nature: imitating the way an eagle, for example, uses its muscles and tendons effectively to grasp and hold prey. To help a drone move more like an animal, Carlone’s team retrofitted it with soft silicone fingers that can pick up and hold objects—a vital skill for machines used in disaster response missions. In a recent study, he says, “We were able to achieve a 92% grasping success rate, showing that the soft gripper enables grasping where a rigid gripper would fail.”
— Christine Thielman
Professor of computer science calls for a new, more systematic approach
Lab draws on advances in computational tools to explore creating proteins from scratch