For years, Zurich, Switzerland has been known as a financial center and home to some of the finest cheeses and chocolates. In recent years the city has been dubbed “The Silicon Valley of Drones”. With both ETH Zurich and the University of Zurich, two of the top universities dedicated to the advancement of drone technologies, it’s no wonder that Zurich has become a hub of drone development in both the research and commercial sectors. Kevin Sartori, Co-founder of the drone software company Auterion said, “Over the past few years, Zurich has become the world’s leading location for the development of the core technology for drones, i.e. software, autopilot, and sensors.” Two such Zurich based startups are UAVenture and Daedalean which have now combined forces to create a new vision based guidance system for drones.

UAVenture was founded in 2014 with a focus in “the development of advanced flight control software and related systems for the next generation of UAV systems.” The software they have developed is called AirRails and is compatible with small to mid sized drones used for mapping, surveying, and surveillance. Though it is ideally suited to such UAVs it is by no means limited to them. Their website goes on to explain that, “The flight controller manages every aspect of the mission autonomously. From takeoff, right through to landing as well as all transitions between muitirotor and fixed-wing flight modes leaving you to focus on the job at hand.”

Daedalean was founded in 2016 by Anna Chernova and Luuk van Dijk with a mission “to build a certified airworthy autopilot that can also pass the exams for human pilots.” As they explain, though there has been tremendous advancement in aviation technology over the last century, the primary way aircraft are flown has not changed. It is still a manual process that relies on human piloting. The challenge is to find a way to pilot an aircraft autonomously while maintaining the necessary safety protocols. In taking on this challenge Daedalean is, “Relying on our expertise in robotics, computer vision and machine learning, as well as a thorough foundation in classical systems engineering, avionics and piloting, we have set out to build guidance navigation and control systems that replace and outperform the human pilot on every measurable scale.”

In 2018 UAVenture and Daedalean announced that they would be bringing together advanced flight control software system and AI autopilot for autonomous flight to create Magpie. Magpie is a drone that can not only fly autonomously, but choose a safe landing spot as well. The drone has a downward facing camera and a mission control computer. The system weighs less 500g and uses less than 15W of energy under normal conditions. Like AirRails, Magpie will be compatible with small to mid sized drones. It will also be fully integrated with AirRails so those already operating on the system will be able to use it right away. Magpie will allow a drone to fly autonomously while avoiding obstacles. The drone will also be able to return to base to land on it’s own, or even land safely away from base.

While in flight the camera takes in the drone’s surroundings and instantly inputs it to the AI computer. The the two devices together work as a pilot would to visually navigate. The computer learns to recognize the environment. It can account for trees, buildings, people, and vehicles. This understanding gives the drone the knowledge of where it is safe to fly. Meanwhile the drone also has sensors that will pick up any obstacles not seen by the camera for added safety. As Simon Wilks, UAVenture’s co-founder and CEO said, “Safety first! Continuous advisory of landing spots during the flight that AirRails gets from Magpie is what makes any UAV mission finally safe for BVLOS in any environment.”

That’s right, safe BVLOS (Beyond Visual Line Of Sight) for any environment. When Magpie has finalized all of it’s test flights, the same ones that would be given to a human pilot and more, it could mean approval for BVLOS. As Magpie learns about it’s environment it will be able to fly and land while removing the risks it could pose to people and property. In a demo video released by UAVenture and Daedalean you can see just how accurately Magpie works. The video shows a drone flying over an open field. An area of grass has been cut back for the drone to land in. Before landing the drone maps the landing zone and separates it into 3 sectors. As the drone is about to land a minivan drives up and parks within one of the 3 sectors the drone had mapped out. The drone instantly takes into account the obstacle and changes it’s flight plan to land safely away from the car. All of this was done with zero human interaction.

Magpie working in tandem with AirRails gives drones the ability to fly with total autonomy, BVLOS, and even if a GPS signal is lost. Luuk van Dijk said of this collaboration, “Magpie is going to be our first widely used product demonstrating the feasibility of autonomous flight control based on neural networks.” As commercial operations become more dependent on drones for a multitude of tasks, Magpie is sure to solve many logistical issues. In fact, Magpie could be the catalyst that pushes drone capabilities to the next level.