Autonomous vehicles are not a novelty, but off-road ones are. Up until now, the only self-driving cars to hit the road have been ones that follow the general rules, recognize traffic signals and lane markings, notice crosswalks and other normal features of a common street. They work solely on the carefully scanned and mapped out well-marked roads. This excludes one-third of America’s public roadways because they are either unpaved or bear no detectable on-road signals like lane markings or stop-here lines… not to mention off-road trails.
Mississippi State University’s Center for Advanced Vehicular Studies took this as an opportunity to create a new kind of autonomous vehicle that could drive on all the roads, take you down any street, and go everywhere and anywhere you want to go. They call it the “Halo Project.” It is a 2014 Subaru Forester modified to be fully electrified and able to drive itself off-road. It made its first debut at this year’s SEMA show (Specialty Equipment Market Association) in Las Vegas.
“The Halo Project itself serves as a platform for our researchers to demonstrate their expertise, on a real-world, high-performance vehicle,” CAVS associate director Matthew Doude said in a statement. “Less than one percent of the Earth is paved, so we needed a vehicle that could be a capable development and test platform both on- and off-road. The Halo Project vehicle is all-wheel drive with tons of wheel torque from its four independent electric motors. This allows us to do research on topics like self-driving cars, even in rugged environments.”
Features of the car include:
- Navigation of a wide variety of terrain autonomously by using a sensor packing that includes LiDAR, radar and cameras. Four LiDAR sensors will create detailed 3D maps of the vehicle’s surroundings. Stereo cameras will serve as the car’s eyes and help it recognize and classify objects. Radar will allow the vehicle to see better through rain and snow, as well as identify types of terrain in front of the vehicle. All of the sensor data will be fed into an onboard supercomputer, provided by NVIDIA.
- Next-generation lithium-ion battery produced by Michigan-based A123, an international leader in battery technology, enabling the vehicle to travel an estimated 230 miles on a single charge. The battery has more than 50 percent more energy capacity than the previous generation.
- Four electric motors, powering each wheel individually and providing over 10,000-newton meters of torque. Built by YASA, the engines are coupled with high-power inverters made by Rinehart Motion Systems.
- Torque to the wheels through custom-designed transmissions provided by Hewland Engineering of Berkshire, United Kingdom.
- Custom bodywork and paint by Clinton Body Shop to ensure that the vehicle looks the part of a supercar. The shop also provided the paint for MSU’s internationally-recognized “Car of the Future” vehicle.
- A suspension that was partially built using steel that was melted, cast and rolled at the CAVS steel research center.
The challenging part of developing an advanced technology such as this is the handling of infrequent or uncommon situations, the events that require performance beyond a systems normal capabilities. Corresponding circumstances on-road might involve: Navigation of construction zones, graffiti on a stop sign (or that looks like a stpp sign), encountering a horse and buggy, etc. Off-road the possible circumstances include the full variety of the natural world, like a tree fallen down across the road, animal crossings (and blockings), flooding and large puddles, etc. This project puts an autonomous car in the most difficult possible scenario: driving in an unknown environment that the car has no prior knowledge of, and with no reliable infrastructure.
The students of MSU’s “Halo Project” took up this challenge by training algorithms to respond to circumstances that almost never happen, that are difficult to predict and that are complex to create. By combining virtual technology with real world experience they created advanced simulations of lifelike outdoor scenes. They use these simulations to train artificial intelligence algorithms to take a camera feed and classify what it sees, labeling trees, sky, open paths and potential obstacles. Then they transfer those algorithms to a purpose-built all-wheel-drive test vehicle and send it out on their dedicated off-road test track, where they can see how their algorithms work and collect more data to feed into the simulations.
“The students and researchers have been working so hard to make this project possible,” said CAVS Executive Director Clay Walden. “I think this car makes an impactful statement about our contribution to the future of autonomy and off-road mobility.” But as far as they’ve come, there is still a long way to go. All in all, what they ultimately hope to achieve with this car is that the technologies they are developing for extreme cases will also help make autonomous vehicles on today’s roads more functional.