Success in robotics competition draws students to LTU program
Work now shifting to artificial intelligence to control cars
The Computer Science Autonomous Robotics Lab at Lawrence Tech University can now be considered a dynasty.
Last week, their team won their sixth consecutive Intelligent Ground Vehicle Competition (IGVC) Self-Drive Challenge held at Oakland University.
The IGVC has two parts. The oldest competition, the auto-nav challenge, has small robot vehicles the size of a kiddie car using various autonomy technologies to find their way around a winding, obstacle-filled course laid out on a parking lot.
In 2017, they added a self-drive challenge for autonomous passenger cars, slightly smaller than a Smart car, which LTU has won every year since its inception.
This year, instead of navigating the car twice around the same course, teams had to complete two different loops around the parking lot — one short loop, one long loop — and various obstacles were moved in and out of the vehicle’s way for it to maneuver around.
LTU beat out Cooper Union in New York, Wayne State, Kettering University in Flint, University of Detroit Mercy and Oklahoma State.
“We had new team members, and they did a great job this year to win again,” said Professor C.J. Chung. “The big improvement was using brain-inspired deep learning technologies in action, such as identifying the fake stop sign and avoiding objects in the road.”
The autonomous car program team is made up of eight computer science students working with two cars and the success of the program has lured new students and alumni alike.
Team captain Justin Dombecki graduated with a bachelor’s degree in computer science in 2008 and returned to earn his masters degree this year.
He went on to work as a mobile app developer for
General Motors after getting his masters degree and has seen the results of what is created at LTU become integrated into the automotive industry.
“Part of what we are doing here is becoming familiar with all the different types of sensors and how to apply them to real life situations,” said Dombecki. “I think a lot of what is going on in the auto industry has come out of the research that is done here.”
What will eventually become part of the auto industry, but is still in development at LTU is artificial intelligence.
AI is being infused into the autonomous cars that Chung and his team are creating for IGVC competitions.
Team member Ryan Kaddis said AI is becoming an important part of their navigation system.
“We are getting more AI, deep learning involved with our vehicle,” he said. “We have been using deep learning for object detection and hopefully for next year we will have deep learning autonomous control.”
Graduate team member Devson Butani is part of a private company team, Snowbotics, that has developed a prototype autonomous snow blower that will soon be going into the manufacturing phase. He said the speed at which LTU students learn and develop technology in their lab has been the key to their success.
“To be able to build these systems from the ground up is what I think is allowing us to win these competitions,” he said.
Helping Chung’s team to create and develop new technology are students from an eight-week National Science Foundation funded summer research program on self-drive technologies.
Micheal Khalfin, a New York native and student at Rice University, Benden Frant, a junior at Harvard from Lake Orion, Shilpi Shaw, a sophomore from Rutgers University and Milan Jostes, a junior at LTU, are part of this year’s program. They are helping to create an “artificial brain” or AI decision making process software to allow a car to complete the competition course.
“I was really interested in getting to work handson with cars,” said Frant, who had been more focused on theoretical computer science. “I can write down code and then see it work in the real world and that is what drew me to this program.”
Jostes said he is looking to be part of the autonomous car team next year.
“LTU does a very good job at giving people the ability to work with real life applications,” he said. “I think the self-driving, autonomous vehicles are a great opportunity to work within a group and create real life scenarios.”
Khalfin admits that AI is an evolving technology that the program and students around the country are trying to harness.
“AI is a bubble right now and every company is putting AI in their description even though they barely use it,” he said. “So we need to test deep learning and see if it is going to work and hack away at it from different angles and even then it still might not work.”
Shaw said what she has learned during her time at LTU will last throughout her college years.
“The algorithms and things that we are building here are definitely things I feel like I could continue developing for a long time and always be able to improve on it,” she said.
Chung said he is hoping to expand the program and add a third car next year. The program and the cars are funded through grant and private funding from companies such as Hyundai Mobis, the National Defense Industrial Association, BAE Systems, MathWorks, Dataspeed Inc., and General Dynamics Land Systems.
“I’ve talked with some sponsors and hopefully I can get an additional vehicle soon for next year’s program,” said Chung. “We want to continue our success.”