MIT Laboratory for Information and Decision Systems

Congratulations to our recent LIDS graduates! Their years of learning and research have paid off and we can’t wait to see what comes next! See a selection of photos from the LIDS Commencement Celebration here and watch MIT’s full ceremony: https://bit.ly/4agLw8K

MIT Schwarzman College of Computing MIT School of Engineering MIT Aeronautics and Astronautics MIT Institute for Data, Systems, and Society MIT EECS Department

PC: Jerard Welcome

Helping robots find the fastest safe route

A new open-source system from MIT and the University of Pennsylvania could help drones navigate complex environments more efficiently.

The trajectory-planning framework, called MIGHTY, allows a UAV to detect and respond to obstacles in milliseconds while maintaining a smooth, efficient flight path. The method generates safe, feasible trajectories faster than leading approaches and is lightweight enough to run in real time using only the robot’s onboard computer and sensors.

The technology could improve autonomous navigation for applications such as disaster recovery, search-and-rescue missions, and package delivery.

The team includes MIT Aeronautics and Astronautics Astro graduate student Kota Kondo, University of Pennsylvania graduate student Yuwei Wu, UPenn professor Vijay Kumar, and MIT professor and LIDS PI Jonathan P. How.

Learn more: https://bit.ly/4e0feBa

MIT Schwarzman College of Computing MIT School of Engineering

Congrats to LIDS PI and Ikigai Labs cofounder Devavrat Shah! Ikigai has just agreed to be acquired by Celonis! Ikigai's decision intelligence, simulation, and forecasting capabilities will now become part of the Celonis Platform. Shah will take on the role of Chief Scientist, Enterprise AI at Celonis, as the entire team continues to build the next generation of Enterprise together. https://bit.ly/4wszdQg

Games people — and machines — play

How do systems make smart decisions when multiple agents, uncertainty, and competing goals are involved?

Gabriele Farina, a LIDS PI and faculty at MIT EECS, combines ideas from game theory, machine learning, optimization, and statistics to better understand strategic reasoning and decision-making in complex environments.

His work is helping advance the foundations of AI systems that must interact, adapt, and make decisions in real-world scenarios.

Read the MIT News profile to learn more: https://bit.ly/4nm4X5G

MIT EECS Department MIT Schwarzman College of Computing MIT School of Engineering

What’s the TRUE cost behind ?

In a recent episode of the India Opportunity Show, LIDS PI Priya Donti speaks with Shrishti Sahu about the growing environmental and economic impact of AI — from energy-hungry data centers to the rapid pace of global AI investment.

The conversation explores whether today’s AI boom is being driven by lasting value or speculation, while also highlighting the technology’s enormous potential to support climate solutions, optimize power grids, and improve energy systems when used responsibly.

The episode also tackles big questions facing countries like India and the global tech community:
Should AI focus on scale or strategy?
How do we balance innovation with sustainability?
What role should policy and corporate accountability play?

Watch: https://bit.ly/4emauGX

MIT EECS Department MIT Schwarzman College of Computing MIT School of Engineering

Exposing The TRUE Cost Behind AI | Ft. Priya Donti The AI boom is happening at an unprecedented pace but are we asking the right questions? In this episode of the India Opportunity Show, Shrishti Sahu sits do...

Inside Efficient AI: From GPUs to GPTs

is powerful — but it’s also incredibly energy-hungry. So how do we make it more efficient?

In the latest episode of Curiosity Unbounded, LIDS Affiliate PI Song Han sits down with MIT President Sally Kornbluth to explore the future of efficient AI — from smarter GPU use to lighter, more efficient models.

Han’s research covers everything from computer vision for autonomous vehicles to improving GPT performance and making image generation more efficient — all with the goal of reducing AI’s growing energy footprint.

He also leads the Efficient AI team at NVIDIA Research, working to optimize next-generation AI systems.

🎧 Listen here: https://bit.ly/3QYPxYI

MIT EECS Department MIT Schwarzman College of Computing MIT School of Engineering

Are we ready for autonomous ride services? 🚗

Before autonomous mobility-on-demand systems become part of everyday life, researchers say we need stronger scientific standards behind the scenes.

A new MIT study published in IEEE Transactions on Robotics, finds that much of today’s research isn’t yet transparent or reproducible enough to guide real-world transportation decisions with confidence.

Improving these standards now could help ensure safer, more reliable, and more equitable systems in the future.

Learn more: https://bit.ly/4cCrAig

The team includes: LIDS grad students Xinling Li and Meshal Alharbi, LIDS PI and CEE Prof Gioele Zardini, DUSP Prof Jinhua Zhao, and collaborators from Stanford University, Google DeepMind, Danmarks Tekniske Universitet - DTU, Technical University of Munich, and ETH Zürich.

MIT EECS Department MIT School of Engineering MIT Schwarzman College of Computing

Before autonomous vehicles scale up, researchers call for stronger scientific standards - cee.mit.edu A new study finds autonomous mobility-on-demand research must become more transparent and reproducible to guide real-world transportation decisions As cities and companies move towards fleets of self-driving taxis, a group of researchers is urging the field [...]Read More...

How do we know if systems are making fair decisions?

As AI is increasingly used to guide decisions in complex systems like power grids, it’s not always clear whether those recommendations treat people and communities fairly.

Researchers from MIT LIDS, AeroAstro, and Saab have developed a new framework—SEED-SET—to help answer that question.

This tool can pinpoint when AI decision-support systems fall short of human-defined ethical standards, making it easier to evaluate fairness across many competing priorities.

Learn more: https://bit.ly/41dEFbr

The team includes LIDS grad students Anjali Parashar and Eric Yang Yu, postdocs Yingke Li and Fei Chen, senior author LIDS PI Chuchu Fan, along with AeroAstro researcher James Neidhoefer and Devesh Upadhyay (Saab).

MIT Schwarzman College of Computing MIT Aeronautics and Astronautics MIT School of Engineering Saab

AI that keeps warehouse robots from getting stuck 🚦🤖

Coordinating hundreds of robots in a busy warehouse is no small task — especially when traffic jams slow everything down.

Researchers from MIT LIDS and Symbotic have developed an system that learns how congestion builds and makes real-time decisions about which robots should go first. By prioritizing robots that are at risk of getting stuck, the system can reroute traffic early and prevent bottlenecks before they happen.

The result? Smoother operations and more efficient warehouses.

The team includes LIDS grad student Han Zheng, postdoc Yining Ma, Symbotic researchers Brandon Araki and Jingkai Chen, and LIDS PI Cathy Wu. The research appears in the "Journal of Artificial Intelligence Research."

Learn more and read the paper: https://bit.ly/4lWdawE

MIT Institute for Data, Systems, and Society MIT School of Engineering MIT Schwarzman College of Computing

🤖 A better way to identify overconfident AI

Large language models can produce answers that sound highly confident — even when they’re incorrect.

New research from LIDS grad student Kimia Hamidieh, PI Marzyeh Ghassemi, and collaborators introduces a more reliable way to measure uncertainty and detect when an AI model is overconfident but wrong.

By helping flag these cases, the technique could make it easier for users to decide when to trust AI predictions — and when to question them.

Learn more and read the paper: https://bit.ly/4sobRZN

The team includes: Kimia Hamidieh, Veronika Thost, Walter Gerych, Mikhail Yurochkin, and Marzyeh Ghassemi

MIT Institute for Data, Systems, and Society MIT EECS Department MIT Schwarzman College of Computing MIT School of Engineering