Robot Swarms in close-proximity flight controlled through Machine Learning

Caltech Engineers have created a new data-driven method of controlling the movement of multiple robots in unmapped and cluttered spaces to avoid collisions with each other. This can be applied to real-life situations such as urban search and rescue to the control of fleets of self-driving cars.

author avatar

01 Dec, 2020

Caltech Engineers have created a new data-driven method of controlling the movement of multiple robots in unmapped and cluttered spaces to avoid collisions with each other. This can be applied to real-life situations such as urban search and rescue to the control of fleets of self-driving cars

The video starts out with 16 Quadrotor drones in equally spaced quadrants. They start their flight on an equal height and distance, then proceed to move into position to start their synchronized movements. In a close proximity flight, the drones are able to mimic the flight pattern of swarms, and as a result were able not to collide with each other. The drones then move back into their initial position and drop down on the mat at the end of the demonstration. 

Soon-Jo Chung, Bren Professor of Aerospace, and Yisong Yue, professor of computing and mathematical sciences, along with Caltech graduates have developed GLAS (Global-to-Local Safe Autonomy Synthesis) as a multi-robot motion-planning algorithm to overcome challenges such as robot coordination in new environments. This algorithm is made to imitate a complete-information planner with existing local information. The have also developed a swarm-tracking controller called "Neural-Swarm" to learn complicated aerodynamic interactions in close-proximity flight. When GLAS and Neural-Swarm are used, the robots learn to navigate through incorporating new information as a “learned model” for movement instead of requiring an initial comprehensive picture of its current environment and the movements of its fellow robots beforehand. 

01 Dec, 2020

I'm a Mechanical Engineer from the Philippines and I've been working for Wevolver as a Content Manager for almost a year now. I'm passionate about innovative technology: 3D Printing, Robotics, Artificial Intelligence (AI), Autonomous Vehicles, etc.

Stay Informed, Be Inspired

Wevolver’s free newsletter delivers the highlights of our award winning articles weekly to your inbox.