Written by: Matthew Busekroos | Produced by: Nate Caldwell
Originally from West Bloomfield, Michigan, Noam Buckman received both his B.S. and M.S. at MIT prior to graduating with his PhD from CSAIL this past February. Buckman credits the exciting projects and cutting-edge research for choosing to spend the entirety of his higher education studies at MIT and CSAIL. Buckman said he was specifically interested to see the progress and innovations in AI and autonomy, especially the Toyota-CSAIL collaboration. He added that he wanted to work on those types of projects in making self-driving cars safer and better at driving on roads with both humans and robots.
“Our lab in CSAIL is very collaborative and supportive,” Buckman said. “When one researcher succeeds, everyone in the lab succeeds and people are always willing to extend their experience and wisdom to improve your research. Daniela and Sertac's style of advising which includes project group meetings encourage this type of osmosis and comradery. In addition, Daniela and Sertac encouraged us to work on problems that will change and impact the world which has motivated the work I focused on.”
During his PhD studies, Buckman worked on algorithms for autonomous vehicles that navigate in environments with humans by considering the internal planning of humans.
“For example, we developed an algorithm using a concept from social psychology known as Social Value Orientation for high-speed driving on highways for ambulances and coordinating vehicles at intersections,” he said. “I also developed algorithms for modeling human driver blind spots and driver failures so that we can develop safer motion planners.”
Additionally, Buckman led the building of the MiniCity which is an experimental testbed for autonomous vehicles and new algorithms.
“In the MiniCity, we can deploy our interactive algorithms on scaled hardware in the presence of other autonomous vehicles and even bring in humans to teleoperate the vehicles,” he said. “The MiniCity has been used as a tool for evaluating neural networks deployed in a city setting, deploying multi-vehicle planning research, and studying human drivers and preferences.”
Buckman said his research will impact industry by showing what is possible if they can explicitly model humans in the presence of robots.
“These algorithms and methods will hopefully be used as a stepping stone (or to inspire) future robotics to create human-aware robots that naturally interact with their surroundings,” he said. “Right now, we are used to robot vacuums that just bump into everything. But we want to move closer to the robots we see in the movies that can empathize with humans and account for the inner decision making of humans.”
Buckman said in the near term, many of these algorithms can be developed for current iterations of autonomous vehicles; and the MiniCity is an example for how scaled robot platforms can accelerate research and development in multi-agent autonomy.
Furthermore, Buckman’s desire for safer, more efficient roads has been a driver of much of his research. He said he dreams of a world without fatal collision thanks to autonomous vehicles.
“In my mind, self-driving cars can provide a solution to this problem; however, there are very difficult challenges that must be overcome,” he said. “For example, we are very good at designing algorithms for doing collision avoidance in static environments or a centrally controlled team of robots. However, roads are full of many actors, pedestrians, human drivers and each person is continuously making decisions that may impact the AV. My research has really been focused on better understanding and accounting for the decisions that other road actors make so that the AV can properly take actions that are safe (and possibly more efficient).”
Since graduating, Buckman has accepted a job at Ford Motor Company. Buckman is a Senior ADAS Software Engineer working on Ford's Level 2/3 autonomy team. He will be helping develop the software deployed on Ford cars for doing things such as hands-free highway driving.