Five years ago, Grayson Randall, a 25-year IBM engineer and holder of nine patents, asked himself, “How hard could it be to design and build a car that can drive itself?” Well, he and his team, Insight Racing, are still trying. And they’re confident they’ve got the right stuff to win this year’s Grand Challenge for autonomous vehicles organized by the U.S. Defense Advanced Research Projects Agency (Darpa).
Starting the team
About seven years ago, Randall mentored a group of high-school students in First (For Inspiration and Recognition of Science and Technology), an international robotic contest aimed at getting kids interested in science and engineering. “The team was quite competitive and earned top honors in the international competition,” Randall says.
Then in 2002, Darpa announced the first of its Grand Challenges in which robotic cars, trucks, even motorcycles would race unmanned over a desert course for prize money and bragging rights. The Challenge would also supply DoD with the latest and greatest in driverless-vehicle technology, part of its move to replace soldiers on the battlefield with autonomous vehicles. Colleagues who knew of Randall and his work in robotics recommended he reprise his role as mentor and team leader and get in the race.
The team began small, with Randall calling in his four former highschool team members, those who had gone on to study engineering at North Carolina State University. Randall thought this would be a great opportunity for his team to learn about robotics on an entirely different scale.
“Initially, it was an exciting way to get students together to work on another challenging project,” Randall says.
The first task at hand was assembling the rest of their team. Fortunately, NCSU proved to be a valuable resource.
“The university helped generate interest in our project and get people involved. For example, we recruited several engineers and students through the school’s robotics initiative,” notes Randall.
The rest of the team was built through connections as the team discovered what kind of engineering backgrounds and specialists they would need. The team eventually grew to around 50 members, all of whom were assigned special tasks in the car’s construction based upon his or her skills. In comparison to other teams, which had full-time staff, Insight’s part-time members certainly didn’t simplify the project.
An Out-of-Pocket Project
The Grand Challenge’s 130-mile desert course presented a particular problem to Randall and his team. Not only did they need to create an autonomous car, but it had to go off-roading as well. Since Insight Racing was self-funded at the time, the team started looking for an affordable vehicle suitable for a desert environment.
“We found an ’87 Chevy Suburban in working condition,” says Randall. “The body was a bit rusty, but the engine was in good shape, so we all chipped in and bought it.”
The Suburban, dubbed the Desert Rat, was retrofitted for autonomous operation based on a design and architecture developed by Randall. However, he kept everyone in the loop by documenting his plans and letting the entire team review them. The team suggested changes to the plans, with all decisions hammered out in group discussions. Soon, individual engineering assignments were handed out.
“Insight Racing is a pretty flat organization with little hierarchy due to the size of our team,” Randall points out. “Each person or small group was responsible for a part of the vehicle. Weekly team meetings let everyone know about the technical and design issues being faced. It was also fabulous from a problem- solving perspective to get the team together and brainstorm because most solutions were better for having the team work together to solve them.”
Although the team solved most problems in group discussions, there were some outstanding issues that had to be decided. It was Randall’s responsibility as team leader to consider all aspects before making the final decision.
Aside from these meetings, team members were kept up to date on design issues, resolutions, and project status via e-mail. Team sponsors and public supporters, on the other hand, got the latest news through regular postings on the team’s Web site.
“One of my tasks as team leader was to monitor stress levels,” Randall says with a laugh. “We all had regular jobs, so all the coding and planning had to be done on our free time. On weekends we came together to install and test the equipment. But there were weekends when something wasn’t working, and I called it off early for the day. It’s important that a team take time off every now and then and have some fun.”
To get the Desert Rat ready, Randall and his team enlisted the donated time, equipment, and engineering skills from the few companies sponsoring Insight Racing. But once the vehicle was ready and the Grand Challenge trials began, the team’s lack of big-time funding began to show.
“We qualified as one of the first 19 teams accepted into the Grand Challenge’s first race in 2003,” says Randall. “But we didn’t have enough money to compete, so we pulled out. We gambled that no one would win and there would be another race.”
And that’s precisely what happened: no vehicle traveled more than 7.5 miles of the 130-mile race, and soon talks of a second Grand Challenge began to circulate.
“We were pretty much selffunded for the second race as well, but because we made the final list for the first Grand Challenge, we had earned some respect and more sponsor interest,” he says. And after the team was selected for the second Grand Challenge’s semifinals, several more sponsors stepped forward.
For their support, sponsors received recognition on the team’s Web site and literature, and key sponsors also got their logo on the Desert Rat.
The Desert Rat ended up placing 12th of 189 participating vehicles. “But if that race had been judged by miles per dollar, we’d have won by a landslide,” Randall says. “We got by on a shoestring budget, while some other teams were spending millions of dollars.”
Switching Gears : From Suburban to Lotus
The next Challenge, dubbed the Urban Challenge, would have vehicles competing in a citylike environment set up in a military base in California. Vehicles would have to respond to traffic laws, road obstructions, and other moving vehicles. So Insight Racing started designing another SUVsized vehicle for it, the kind of vehicle that would give the team the room to carry the equipment it would need.
Then a lucky meeting brought together Randall and executives from Lotus, the British car company. They had heard of the Grand Challenge and were interested in participating, but they had not found the right opportunity.
“We talked about the new Challenge, and soon Lotus offered us an Elise sports car to use as our platform,” says Randall.
“Our initial response was, it’s an awfully small car,” he recalls. To switch to the Elise, the team would have to miniaturize the car’s electronics. “But this miniaturization was a great opportunity to take a step toward accommodating our technology for commercial vehicles. So once the team decided it was a challenge we were willing to take, we worked with Lotus to put the plans together.”
Miniaturization required distributing the car’s computers, which had previously sat in a 19-in. rack in the back of the Desert Rat. Now the team needed to find nine computers that would fit in a considerably more compact space, yet still have the power, electrical characteristics, and bandwidth needed to store and process the huge amount of information from the car’s sensors. The equipment also had to keep cool, despite close quarters near the rear-mounted engine.
Lotus quickly became an invaluable team player, solving the power and cooling problems and providing technical support throughout the car’s construction. For example, they converted the little blue Elise to drive-by-wire technology and helped install the much-needed air conditioning.
To navigate, the car needed “eyes.” So Insight Racing added several different types of sensors to the roof, front, and rear of the vehicle. Sick Inc., a sponsor headquartered in Minneapolis, supplied Laser Measurement System (LMS) sensors for the front and rear bumpers. “The LMS uses a 2D infrared laser beam, which is deflected by two rotating mirrors. This creates a 2D sweep, and that sweep provides distance information,” says Jeff Wuendry, product marketing manager for Sick. The company also provides sensing equipment and technical support to two-thirds of the teams participating in Darpa’s Urban Challenge. Wuendry explains that the technology Sick provides to each team is identical, but that each team uses their skills and creativity to make each application their own.
Another key component came from sponsor Comtrol Corp., Maple Grove, Minn. They donated their DeviceMaster technology and support. “DeviceMaster collects data from the car’s GPS units, cameras, LMS, and other sensors, and sends it to the car’s computer, letting the sports car make decisions in real-time,” says Kris Rydberg, director of product development and corporate strategy for Comtrol. “Data flows at 500 kbps, so decisions take only milliseconds. From a technical perspective, not much technology can perform at that speed.”
Insight Racing believes it has miniaturized all the technology needed for the Urban Challenge. But, as Randall and his team know, an engineer’s work is never done.
In the few weeks before the race, the team still needs to iron out a few problems, such as making sure the car acts appropriately at intersections and can negotiate its way through a parking lot, all tasks in the upcoming Urban Challenge. “We’ve essentially solved the data problems for this race, now that we have sufficient network bandwidth and computing power. But we still need to teach the vehicle to be smart enough to do all the maneuvers required,” Randall admits.
The Insight team is pleased with the Lone Wolf’s design and implementation, but the technology will have to evolve further to reduce costs and size if it is to ever be commercially offered. The team’s grassroots and low budget may give them a leg up in this area. “Compared to their competitors, Insight has kept their project economical,” says Comtrol’s Rydberg. “So when the government asks the teams, ‘Have you considered space restraints and costs?’ Insight can say that they definitely have.”
For now, however, Insight Racing team’s focus is on the Urban Challenge, and for them, the race is on.
Tips for managing an engineering team
Grayson Randall has learned quite a bit about engineering teams and what it takes to make one work. Here are a few management tips he has found particularly useful:
1. Do something you are passionate about.
“It’s important that team members be passionate about working on a project and are willing to work hard,” Randall says. “With all the hardware and software challenges, you have to keep that passion and excitement.”
2. Solve problems one step at a time.
“It’s important that your team not get overwhelmed, and engineering problems can be very overwhelming at first,” says Randall. “As team leader, set achievable goals, starting with just a piece of the problem and solving it. Then tackle another piece of the problem and solve that. Pretty soon, the big problem is solved.”
3. Concentrate on what’s most important.
“Teams often get stuck trying to solve a particular problem and lose track of the overall goal. So there are times you need to put certain problems aside and come back to them later,” notes Randall.
4. Make sure your team has fun.
“It’s important for the team to be able to laugh and not get too stressed out when things go wrong. I’m lucky that our team always has a good time. Sponsors joke that we have too good of a time,” he says. “Then when it’s time to work, everyone kicks in and enjoys what they’re doing.”
5. Know the importance of distributed development.
“From a development process perspective, everyone needs to work on their own time. Then on weekends we get together and put our technology to the test,” Randall says. “Because everybody has another job, we use e-mail and weekly meetings to get business strategies together. With these restraints, it’s important that everyone work on their assigned components of the project and resolve issues on their own time.”