Watch our OverDrive autonomy stack pilot a Polaris RZR through dense vegetation and rutted creek beds before breaking out to race across an open field. This type of terrain is challenging: the tall vegetation obscures the complex ground geometry and presents difficult choices for the autonomy stack. Which vegetation can be pushed through and which should be avoided? How should the vehicle approach rough ground when it is partially occluded by tall grass and bushes? OverDrive's onboard AI leverages years of data collected through continuous testing on a fleet of vehicles in a wide range of biomes to make these difficult decisions, allowing it to navigate even the most challenging terrain reliably. This is real ground autonomy: (1x) video of high-speed, point-to-point navigation powered by onboard sensing and compute.
This video could have been made just by choosing a known good GPS route to follow and simply ignoring all the sensors. How do we know this is actually making decisions about which vegetation to push through? What sensors are you using to determine the density of the vegetation? How do you know the difference between a soft weed and a soft weed that’s hiding a tree stump?
This is part one of my ideal, real world application- mountain bike shuttle. Hear me out- you drive up the mountain, grab your bike off the rack and take the trail down while your razor drives itself back down to meet you at the trailhead.
Invasive noxious weeds is a real problem in far north Queensland, beyond manual control. This technology is part of the solution. Intelligent autonomous ground vehicles augmented by satellite imaging, low level drone imaging, plant identification, and finally herbicide management.
Looks like just out of a video game!
Now this is some serious off-road AI! Navigating through dense vegetation and rough terrain in real-time is no joke, but OverDrive makes it look effortless. Crazy to think about the split-second decisions it’s making on the fly.