Stanford Racing Team
Sebastian Thrun, Associate Professor, Mike Montemerlo, Technical Director, Stanford Racing Team, talks about the qualifying event for Stanford’s autonomous vehicle “Junior”, in Mountain View, California. 53 teams in total will compete, this being one of Darpa’s first stops along the way.
Junior is a 2006 Passat wagon whose steering, throttle, and brakes have all been modified by engineers at the Volkswagen of America Electronics Research Lab in Palo Alto, Calif., to be completely computer-controllable. The engineers also have created custom mountings for a bevy of sophisticated sensors.
An important difference between Junior and Stanley is that Junior must be aware of fast-moving objects all around it, while Stanley only had to grapple with still objects in front of it. Juniors sensors are therefore much more sophisticated, Thrun says. They include a range-finding laser array that spins to provide a 360-degree, three-dimensional view of the surrounding environment in near real-time. The laser array is accompanied by a device with six video cameras that “see” all around the car. Junior also uses bumper-mounted lasers, radar, Global Positioning System receivers, and inertial navigation hardware to collect data about where it is and what is around.
Because Junior collects much more data than Stanley did, it’s computational hardware must be commensurately more powerful, says Montemerlo. Using Intel Core 2 Duo processors – each chip includes multiple processing units – Junior’s “brain” is about four times more powerful than Stanleys.
But what makes Junior truly autonomous will be its software, which is the focus of about a dozen students, faculty, and researchers at the SAIL. Modules for tasks such as perception, mapping, and planning give Junior the machine-learning ability to improve it’s driving and to convert raw sensor data into a cohesive understanding of its situation.
New software development began last fall. Montemerlo has been testing some of the team’s software modules in simulated traffic situations since the beginning of the year. The team expects to move into full-time testing and iterative improvement by the end of March.
Junior’s name is not only an implicit homage to its predecessor, but also to Stanford University’s namesake, Leland Stanford, Jr. Carrying this sense of history, Junior will set out to make a technology history of its own and pave the way to a future where autonomous cars can make driving safer, more accessible, and more efficient. Self-driving cars could give drivers newfound free time.
Sebastian Thrun Stanford Robot department: And this is the 2007 Urban Challenge. Now, why are we doing this? Well, as we all know, cars are unsafe. We drive cars mainly today, and in the process of doing so, we kill something like forty-two thousand people every year on U.S. highways, and about a million worldwide, and it’s largely because of human error.
So making cars safer is a chief concern to us as a society and to us at Stanford. Cars are inefficient; they are fuel inefficient, they make inefficient use of highways, and they make inefficient use of our time. Most of us spend spent more than an hour per day in commute traffic, and that something that we shouldn’t spend attention on, which is just driving in traffic jams like this morning on 280.
So the idea of self-driving cars is a really big idea that will really change society in my opinion. Now, the 2005 Grand Challenge, that we happened to win, was the first step. It was a great step in driving along a desert road, but it was deficient in many ways. A desert road is clearly not representative of the traffic situation we encounter every day. So in response to those deficiencies, DARPA has created the Urban Challenge, which is driving in a city, and this is an event that leads up to a big race going to take place November Third somewhere in California, we don’t know where, called the Urban Challenge, where on the order of twenty vehicles will race in a city, and basically run errands. Like a Fed Ex delivery truck, for example, going from here to here.
And in the process of doing this, they’re going to encounter other vehicles, human-driven, and robotic vehicles, and they have to be able to negotiate these situations. That to us is a huge technical challenge, so we’ve focused on the technical side. We have to build cars that can understand the world, that can understand other vehicles. Sense them, perceive them, make predictions, and interact with them. So you have to understand that if you come to a stop sign, that the other car might be there first, it might want to go first, there are certain rules that govern regular traffic, and you have to be able to adhere to these rules.
You have to make machines that replicate human thinking in that specific domain. It turned out to be a major challenge. It’s been done, worked on for the last two decades in the United States and elsewhere, and this is the first event really testing autonomous vehicles driving in urban situations. I think it’s the next step in a number of steps that will eventually lead us to self-driving cars and safer cars. With that, I welcome you all and hand over to Mike Montemerlo who has been the technical director this year.
Mike Montemerlo Stanford Roadrunner team: I just want to give you a brief outline, for those of you who aren’t familiar with what the site visit is going to look like today. The first step is just taking care of some paperwork that we have to demonstrate the citizenship of the team leader, and everybody has his driver’s license ready to go. Then the first thing is to demonstrate the safety of the robot.
So we have an emergency kill system, that we can wirelessly pause or disable the car from outside the car. We had to demonstrate this to DARPA, that we can bring the car to a complete, immediate stop after – at a speed of twenty miles per hour. You’ll see it’s going to start over, in the far corner of the parking lot, and then come down towards our simulated intersection here, and then when it reaches twenty miles per hour they’re going to pause the robot and make sure it stops within a fixed amount of, within a minimum amount of distance. The next step is, that the actual, the real test, and there are two parts to it. There’s a navigation test, and a traffic test. Navigation tests the basic behaviors of the robot. So, can it stay in its lane, does it stop for stop signs, can it avoid static obstacles, can it do u-turns, and can it basically follow the missions that DARPA provides you with?
As Sebastian said, these missions consist of going from checkpoint one to checkpoint five to checkpoint nine, for example. And during that test, there won’t be anyone in the car, so they’ll be following it with a chase car behind, and if something goes wrong they can always hit the pause or the kill button. And the traffic test involves more complicated traffic situations, where there are actually two human-driven cars on the course at the same time. And you’ll see there’s going to be a gray Touareg, and thereÃs also a black sedan on the course, driven by Stanford Racing Team people.
They’re going to test more complicated behaviors. For example, intersection precedence is on that we expect to see. So that Junior wait’s until it’s his turn at the intersection to go. And also, there’re stopped, stopped cars, Junior will be expected to come to a safe stop in front of the car, look to see if there’s anybody in the other lane, and actually cross the double yellow line and pass the parked car. That’s about all of it.