Lex Fridman
fridman@mit.edu
I'm a research scientist at MIT, working on human-centered artificial intelligence. In particular, I'm interested in developing deep learning approaches for perception, planning, and human-robot interaction in the context of real-world shared autonomy systems.
Recent News
- Received Best Paper Honorable Mention Award at CHI 2018 for our work on cognitive load estimation in video.
- Launched Artificial Intelligence podcast.
- Taught: 6.S094: Deep Learning for Self-Driving Cars
- Explored the nature of intelligence: 6.S099: Artificial General Intelligence
- Received Best Paper Award at CHI 2017 for our work on driver glances.
MIT 6.S094: Deep Learning for Self-Driving Cars
A course on the practice of deep learning explored through the theme of building a self-driving car. Course page is http://selfdrivingcars.mit.edu. Besides lectures and guest talks, it included a deep reinforcement learning competition (DeepTraffic) and an end-to-end driving simulation (DeepTesla).
- Lecture 1: Introduction to Deep Learning and Self-Driving Cars
- Lecture 2: Deep Reinforcement Learning for Motion Planning
- Lecture 3: Convolutional Neural Networks for End-to-End Learning of the Driving Task
- Lecture 4: Recurrent Neural Networks for Steering through Time
- Lecture 5: Deep Learning for Human-Centered Semi-Autonomous Vehicles
Select Videos: Research, Demos, and Talks
We ran our new glance classification code on the driver-facing video from the tragic Uber crash in Arizona. Open source code and arXiv paper will be out soon. I hope it can help, in however small a way, make the roads safer for people inside and outside both manually-controlled and autonomous vehicles.
MIT-AVT study aims to understand, through large-scale real-world driving data collection and large-scale deep learning based parsing of that data, how human-AI interaction in driving can be safe and enjoyable. See website for more technical details (including paper).
A demonstration of our gaze region classification algorithm on five synchronized video streams in a Tesla while transfering control to Autopilot and taking control back.
We present a method for detecting driver frustration from both video and audio streams captured during the driver's interaction with an in-vehicle voice-based navigation system.
This is a demo for our automated synchronization algorithm for driving data that uses vibration and steering events to sync driving data within 13ms precision.
Publications
MIT Autonomous Vehicle Technology Study: Large-Scale Deep Learning Based Analysis of Driver Behavior and Interaction with Automation
Journal: IEEE Access
Arguing Machines: Perception-Control System Redundancy and Edge Case Discovery in Real-World Autonomous Driving
Conference: AAAI
DeepTraffic: Driving Fast through Dense Traffic with Deep Reinforcement Learning
2018
Under Review
PDF available on request
Conference: AAAI
Designing Toward Minimalism in Modern Vehicle HMI
2018
Under Review
PDF available on request
Conference: CHI
Road Curvature with Tesla Autopilot: Difference in Approach between Human and Machine
2018
Under Review
PDF available on request
Conference: CHI
Cognitive Load Estimation in the Wild
2018
Under Review
PDF available on request
Conference: CHI
Understanding Non-Verbal Communication Between Vehicles and Pedestrians Toward Safer Automated Vehicles
2017
PDF available on request
Conference: FAST-Zero
SideEye: A Generative Neural Network Based Simulator of Human Peripheral Vision
Conference: CHI
Owl and Lizard: Patterns of Head Pose
and Eye Pose in Driver Gaze Classification
and Eye Pose in Driver Gaze Classification
2016
PDF
Journal: IET Computer Vision
Automated Synchronization of Driving Data
Using Vibration and Steering Events
Using Vibration and Steering Events
2016
PDF
Journal: Pattern Recognition Letters
Driver Gaze Region Estimation without Use of Eye Movement
2016
PDF
Journal: IEEE Intelligent Systems
Behavioral Impact of Drivers' Roles in Automated Driving
2016
PDF available on request
Conference: AutomotiveUI
Distinguishing Patterns in Drivers' Visual Attention Allocation
Using Hidden Markov Models
Using Hidden Markov Models
2016
PDF
Journal: Transportation Research Part F
Multi-modal Decision Fusion for Continuous Authentication
2015
PDF
Journal: Computers and Electrical Engineering
Active Authentication on Mobile Devices
via Stylometry, Application Usage, Web Browsing, and GPS Location
via Stylometry, Application Usage, Web Browsing, and GPS Location
2015
PDF
Journal: IEEE Systems Journal
On the Joint Impact of Bias and Power Control on Downlink Spectral Efficiency in Cellular Networks
Journal: IEEE/ACM Transactions on Networking
Active Linguistic Authentication Using Real-Time
Stylometric Evaluation for Multi-Modal Decision Fusion
Stylometric Evaluation for Multi-Modal Decision Fusion
2014
PDF available on request
Conference: Advances in Digital Forensics
Cross-Layer Multicommodity Capacity Expansion
on Ad Hoc Wireless Networks of Cognitive Radios
on Ad Hoc Wireless Networks of Cognitive Radios
2008
PDF
Conference: CISS
A Simplification Algorithm for Visualizing
the Structure of Complex Graphs
the Structure of Complex Graphs
2008
PDF
Conference: International Conference on Information Visualisation
Distributed Path Planning for Connectivity Under Uncertainty
by Ant Colony Optimization
by Ant Colony Optimization
2008
PDF
Conference: American Control Conference (ACC)