Seth Hollar is an engineering professor and entrepreneur who is involved—fittingly enough—with North Carolina State’s Engineering Entrepreneurship Program. He works with college students who want to start their own companies.
Their current focus is EcoPRT. The budding company is building an economical, ecologically friendly personal rapid transit system. In other words, they’re building really small cars that can move students around NC State’s campus—like Ubers, but with just enough room for two people.
Except these cars drive themselves.
We visited the EcoPRT lab to see the vehicles and learn about the technology.
Tell us about personal rapid transit and EcoPRT.
That concept has been around since the ’70s, more or less. The whole idea is that they were relatively small vehicles that were automated on their own roadways or dedicated roadways.
We’ve added additional autonomy where the vehicles can navigate on existing roads or pathways, in addition to being on their own dedicated road ways.
They’re very small: they only hold two people and weigh just five hundred pounds.
What’s your long-term vision for this technology?
For a growing area like Raleigh [where NC State is located], as people move in here, traffic gets worse and worse. Typically only one or two people ride in a vehicle at a time. So you have a three- to five-thousand pound vehicle moving one person around.
We’ve trying to re-envision what a roadway could be. For us, that is something that could be no more than three or four feet wide. Instead of taking up a space that might be on the order of 150 square feet, this might be on the order of 30 square feet.
This is not a high speed vehicle. It’s designed to move people within a city.
You have students helping with the project. What subjects do they study?
We’ve had civil engineer students and design students, but the bulk of our students are electrical, computer, and mechanical engineering students. We’ve also had computer science students.
The project is fairly large in scope. You have to design and build the mechanical structure of the vehicle. You have to coordinate a group of vehicles together. You have a number of sensors and the processing navigation on the electronics that fall more on the electrical and computer engineer side.
What are the challenges on the mechanical engineering side?
One of our vehicles is fully built and operational. Another is nearly complete.
When you put the shell on both these vehicles, they will look identical, but based on lessons learned with the first one, we redid some of the design.
For instance, we looked the suspension. Being a narrow vehicle, it can be a little top heavy. If you don’t have the suspension right, then ride experience can be a little bit like a boat: it can rock back and forth. You don’t want the vehicle to bend over too much as you make turns. So, we looked at how can we design the suspension to better handle that.
What about the challenges on the electrical side?
We decided early on to use a gaming laptop to operate it, just because of the ease of obtaining a laptop. But a lot of the interfaces had to be created to connect with that laptop.
We have something called linear actuators that operate the breaks. Those breaks have to somehow communicate to the laptop, so we’ve designed our own printed circuit boards that drive the actuators and sense the motion. That’s done for the steering, too. We also use the interface with the motor controller.
That’s just the hardware side. On the software side, you have to actually program that. How do you get it to follow a path autonomously? Then you start thinking about what type of sensors do you need to interface with the vehicle so that it can know where it is.
There are a number of different sensors. There is a inertial measurement unit that measures acceleration and rotations rates. That gets combined with the GPS to tell you where you are on a map. Then we add code to create a navigation algorithm so it can follow a particular path.
The other element is, what if there’s something in the way in the path? We use two main technologies to detect objects. One is a LIDAR technology, which uses lasers to create a 3-D image of the environment.
We also use cameras. People have two eyes, which helps us determine depth. Like two eyes on people, we use a stereo camera that has two camera units that can also determine depth.
The task with these sensors is you have to grab the data, process it, associate it with how the vehicle is navigating, and make high-level logic decisions about what exactly to do.
There are a variety of different challenges and steps to get everything in place and working. How do we process an image to extract depth measurements or to classify an object as something the vehicle recognizes? How do we fuse LIDAR and camera data together? How do you get the vehicle to navigate the obstacle?
So your plan is to have a number of these vehicles working together. What challenges arise from that?
There are a couple different aspects to that.
There is a side where the server needs to communicate to the vehicles, and the vehicles need to communicate to it. So the vehicles are constantly sending data about where they are and any basic status information that they have.
The server needs to provide a path for the vehicles to travel to their next destination. The server has to be smart enough so that it doesn’t create conflicts—mainly you can’t ask too many vehicles to follow a certain path, and you can’t ask two vehicles to follow the same path simultaneously, because you risk them colliding.
The other side of it is, you might ask to go from point A to point B. The server knows that, and the server knows the thirty or forty or fifty other students who are also asking for routes, so it has to be smart enough to know how to meet demand from students after following the route from A to B. Once it drops these people off, it needs to figure out how to replace the vehicles so that when somebody calls for it, it doesn’t take very long for a vehicle to come.
How close are you to having these vehicles out taking people across campus?
We have a three-phase approach.
The first phase is just working out the technical details. We have a parking lot out there where we can do some testing. The next phase is operating in a different environment, still with just researchers for testing purposes. The third phase is doing short runs with actual people on campus, but in a controlled environment.
Once that third phase is successful, we hope to add more vehicles and expand the route and duration at which they run. But we’re still kind of early on in the three-phase approach.