After a journey that involved a crash, a dead battery, and late-night coding sprints, a Virginia Tech engineering team took home a third place finish in the first year of an autonomous vehicle competition held by Society of Automotive Engineers (SAE) and General Motors.
Student-led Victor Tango AutoDrive was the sole team — out of eight total collegiate design teams from the U.S. and Canada — to successfully complete all three portions of the AutoDrive Challenge, held at General Motors Desert Proving Ground in Yuma, Arizona.
The challenges were based on different aspects of driving that are basic for humans, but require complex perception, navigation, and behavior algorithms for a self-driving car, such as stopping at stop signs or staying in lane lines.
Currently, fully autonomous cars do not yet exist. By hosting the AutoDrive competition, General Motors and SAE are further catalyzing development of technology that may one day lead to entirely self-driving cars — a future possibility that can only be achieved with intensive research and rigorous testing.
The road to Yuma
For Victor Tango AutoDrive Team, the road to Yuma itself was the first challenge.
“It is really an underdog success story,” said Andy Cohen, a senior from Crozet, Virginia, studying mechanical engineering and a member of the controls subteam and the business subteam lead.
Weeks prior to competition, the team was modifying their competition-provided 2017 Chevrolet Bolt electric vehicle when a member of the team pressed the wrong pedal while switching out of drive and accidentally crashed the vehicle into the garage door of their testing lab, located in a parking garage near the University Gateway Center.
“We weren’t sure if it was going to be totaled or whether the $100,000 of hardware we had mounted on the roof was destroyed,” Cohen said. “This was a really scary moment for us. This was our worst nightmare imaginable.”
They assessed the damage and declared the incident “a minor setback,” Cohen said. With a few fixes, they were back up and running in a week’s time.
Then the battery died. They tried charging the electric vehicle at stations all over campus and a Chevrolet dealership, and found that not only had it died, it would no longer charge. With three weeks left to competition and local dealerships and specialists unsure how to fix the car, their only option was to send it back to General Motors in Detroit, Michigan.
The team was left to wait. With no information on the status of their car, the team had to make an executive decision — they were not going to be able to compete. They sent a letter to General Motors citing safety risks — even if they got their car back in time, they would need dozens of hours of testing to ensure the safety of the students operating the vehicle from inside.
Despite the hours, stress, and late nights already poured into the competition, the team didn’t think it could be done, and the window closed on the first year of the AutoDrive competition.
“The next half-week was probably the most blissful experience we’ve had of our lives,” Cohen said, reminiscing on the newfound time team members had. “It was almost a bittersweet feeling when, later that week, we got this out-of-the-blue notice from [General Motors] and they say, ‘We’ve fixed your car, we’re sending it back tomorrow.’”
So they made another executive decision: If they were going to do this with 10 days left before competition, they were going to have to complete 100 hours of testing.
“It was a total rollercoaster of emotions,” Cohen said.
They didn’t have time to process — they just got to work. They coded all night and into the following mornings, completing all of the tests they set out to accomplish, guided by project manager Emma Tum Suden, a senior from Orlando, Florida, studying mechanical engineering.
Ironically, Cohen said, it was this time crunch that enabled their success in the competition. Rather than go with their first plan using complicated methodologies for keeping in lane lines, they went with a “fusion of old school and new school,” integrating traditional environment mapping with complex perception algorithms.
“Everyone else at the competition hadn’t relied upon those maps because they figured they could do it entirely with perception,” Cohen said. “But what was our saving grace was that the environment was not nearly as well regulated as the other teams thought it was going to be.”
Because their vehicle could navigate the faded lane lines, Victor Tango AutoDrive Team was the only one to complete the second, lateral challenge.
It helped, Cohen said, that they started off with a particularly inspiring moment during the first challenge: As they watched their vehicle successfully navigate the stop sign challenge on the second try, they blasted "Enter Sandman" on Cohen’s phone speaker.
“That really set the tone for the rest of the competition,” Cohen said.
Structure of the competition
Each team was tasked with using a 2017 Chevrolet Bolt electric vehicle as a platform to develop a fully autonomous passenger vehicle. In Arizona, teams presented on the social impacts of autonomous vehicles then put their electrical designs and coding to the test by navigating a closed test track in autonomous mode.
The competition is a three-year-long process. During this first year, teams focused on concept selection by becoming familiar with sensing and computing software. They were tasked with completion of a written concept design paper and an on-site evaluation that included dynamic tasks, including straight roadway driving and object avoidance and detection.
The Victor Tango AutoDrive Team placed in the top three out of eight total teams in each category:
Social Responsibility Report – 3rd
Social Responsibility Presentation – 3rd
Mapping Challenge – 3rd
Lateral Challenge – 1st
Object Detection and Avoidance Challenge – 3rd
Each year the competition gets progressively harder. The car will eventually have to carry out more complex tasks, such as moving at higher speeds, making U-turns, and dodging a dynamic object.
The hope is that by 2020, the AutoDrive team will be able to produce a level four autonomous vehicle. According to SAE standards, level four autonomy means the vehicle will be able to complete all critical driving functions such as steering, braking, and responding to traffic without the need for human intervention — to the point where a driver could take a nap while in transit.
SAE standards identify level five autonomy as the final level. A level five autonomous vehicle can do all that a level four vehicle can do, but can do so even in extreme and unconventional environments, like a dirt or unlined road.
About the team
This isn’t the first iteration of the Victor Tango team: 11 years ago, a Virginia Tech team called Victor Tango began working on autonomous vehicle technology and competed in the Defense Advanced Research Projects Agency (DARPA) Urban Challenge.
The autonomous vehicle the first team created was one of three vehicles in the competition to successfully traverse across a 60-mile terrain through challenging obstacles, aided by a partnership between Virginia Tech and Blacksburg-based, Hokie-run company TORC Robotics that began in 2005. Together, they have been improving autonomous vehicle technology since.
Building on what was started by Victor Tango, professors in the College of Engineering helped launch the Virginia Tech AutoDrive team in August 2017.
Al Wicks, associate professor of mechanical engineering, along with several other faculty members, pitched the team to SAE. Because of Virginia Tech’s previous success in the DARPA competitions, the team was selected to compete.
Victor Tango AutoDrive Team is composed of dozens of engineering, computer science, and business students and faculty from the Department of Mechanical Engineering, the Department of Computer Science, the Charles E. Via Jr. Department of Civil and Environmental Engineering, the Bradley Department of Electrical and Computer Engineering, and the Pamplin College of Business.
The AutoDrive competition was sponsored by General Motors and SAE, with additional support from Continental, Intel, MathWorks, and Velodyne LiDAR. Victor Tango AutoDrive Team received support during the first year of competition from TORC Robotics, Odoo, Edmund Optics, and the Virginia Tech Transportation Institute.
Written by Erica Corder and excerpts written by Stephanie Kapllani