AI drone flies through dense forest at 25mph

Until recently, only well-trained FPV drone pilots were able to fly at high speeds in thick forests. However, owing to advances in artificial intelligence research, scientists have created an AI drone that can fly at 25 mph through a dense forest while avoiding any branches and other obstacles. The AI drone is unique in that it does not need to be taught in a comparable setting.

AI drone flies through dense forest at 25mph

The barrier is data processing.

The operation of the autonomous flying drone was recently published in the Scientific Journal Science Robotics. “Quadrotor drones are very agile of themselves, but so far only human drone pilots were able to get the extreme out of the drone,” the researchers write. “Autonomous flying based on built-in sensors and computing power was so far limited to low speeds.”

The multiple tasks of observation, mapping, and planning are accomplished sequentially by the current generation of autonomous flying drones that operate on the basis of Artificial Intelligence (AI). In other words, the AI drone first determines the surroundings, then converts the environment into a 3D representation, and then determines the next piece of the flight path. However, in more complicated situations such as a dense forest, such a strategy is both too slow and too inaccurate. As a result, the bottleneck is the method through which the data is processed.

A novel AI technique

A new kind of computer vision and artificial intelligence (AI) can alter this, according to researchers at the UZH Robotics and Perception Group. According to the synopsis of a newly released research paper:

The key principle is to directly map noisy sensory observations to collision-free trajectories in a receding-horizon fashion. This direct mapping drastically reduces processing latency and increases robustness to noisy and incomplete perception. The sensorimotor mapping is performed by a convolutional network that is trained exclusively in simulation via privileged learning: imitating an expert with access to privileged information.

By simulating realistic sensor noise, our approach achieves zero-shot transfer from simulation to challenging real-world environments that were never experienced during training: dense forests, snow-covered terrain, derailed trains, and collapsed buildings. Our work demonstrates that end-to-end policies trained in simulation enable high-speed autonomous flight through challenging environments, outperforming traditional obstacle avoidance pipelines.

AI drone takes after a well-trained FPV drone pilot

In other words, the drone can autonomously decide a path through an area that is wholly unfamiliar to him, without previously being taught in the circumstance at hand. And all at a rapid rate, owing to the shorter data processing pipeline. The AI drone is modeled after a skilled FPV drone operator. The following video demonstrates the new technology:

Applications

Drones will be able to be deployed autonomously by all kinds of complicated surroundings in the future, according to the algorithm. Consider Search and Rescue applications in a difficult-to-access or partially collapsed location. Of course, there are the appropriate military applications to consider.

What do you think about AI drones such as the one in this article? Let us know in the comments below.

Get your Part 107 Certificate

Pass the test and take to the skies with the Pilot Institute. We have helped thousands of people become airplane and commercial drone pilots. Our courses are designed by industry experts to help you pass FAA tests and achieve your dreams.

Stay in touch!

If you'd like to stay up to date with all the latest drone News, scoops, rumors, and reviews follow us on Twitter, Facebook, YouTube, Instagram, or… Subscribe to our Daily Drone News email.*

FTC: DroneXL.co uses affiliate links that generate income. * We do not sell, share, rent out or spam your email, ever. Our email goes out on weekdays around 5:30 p.m.

This article first appeared on Dronewatch and is written by Wiebe de Jager who is also a DroneXL contributor.