Autonomous Drifting

From AMREL:

You know how the stuntmen make fast cars drift in action movies? Have you ever wanted to make a remote-controlled toy car drift like that? Of course you have.  If there ever were awards for endeavors that sound silly, but is actually technically interesting, then the folks at MIT’s Aerospace Controls Lab would surely be nominated.

Unmanned systems are rarely fully autonomous.  Instead, researchers are pursuing “sliding” autonomy, i.e. an operator retains control, while some behaviors are made autonomous. Aerospace Controls Lab decided to teach a remote-control toy car how to autonomously drift.

They started by running their learning algorithm through simulations.  Information from these simulations was transferred to performance modifiers. When the car was run through its drifting actions in reality, the algorithm was constantly modified. The result is a car that can maintain drifting in a full circle even when salt is added to the floor, or another vehicle interferes with it.

 

Featured Product

Boston Dynamics Webinar - Why Humanoids Are the Future of Manufacturing

Boston Dynamics Webinar - Why Humanoids Are the Future of Manufacturing

Join us November 18th for this Webinar as we reflect on what we've learned by observing factory floors, and why we've grown convinced that chasing generalization in manipulation—both in hardware and behavior—isn't just interesting, but necessary. We'll discuss AI research threads we're exploring at Boston Dynamics to push this mission forward, and highlight opportunities our field should collectively invest more in to turn the humanoid vision, and the reinvention of manufacturing, into a practical, economically viable product.