Russian businessman Oleg Kupervasser left for Israel at 25 in 1991; he later returned to Russia to establish a startup at Skolkovo in 2012 to develop a video-based navigation system for unmanned aerial vehicles. However, he failed to raise enough funding for the project at Skolkovo, so he went to Israel again to attract investment, he told Invest Foresight.
From Technion to Skolkovo and Hangzhou
Before starting his own project, Oleg Kupervasser worked at Technion (Israel Institute of Technology) on a computer vision system for Rafael rockets. The company is one of the four largest manufacturers of military equipment in Israel.
“When developing computer vision-based navigation we are actually creating a robot; its ‘eyes’ are not even the most important part of the system – it is its brain,” the scientist says. “We see with our brain, not just our eyes. It is the brain that creates a three-dimensional visual image.”
At Skolkovo, Kupervasser wanted to find investors for his project and go into production. He set up the Transist Video company there, but all he could raise in Russia was a RUR 1 mio ($15K) grant from the Foundation for Assistance to Small Innovative Enterprises in Science and Technology (the Bortnik Foundation), which supported the startup for a year. In 2014, the Chinese GPS navigator manufacturer Hangzhou Avisi Electronics found Transist Video on the Skolkovo website, and proposed joining its video-based navigation project and helping with finding investment in China. In 2017–2018, Avisi and the Chinese Communist Party provided $125K to Kupervasser, who wrote a video-based navigation program and tested it on a real drone.
Why do we need video-based navigation?
The most common navigation services are satellite (GPS, Glonass in Russia) and inertial navigation (accelerometers, gyroscopes) Over time, inertial navigation accumulates errors and needs to be complemented with satellite devices. Then the number errors stops growing although sometimes satellite signal is weak due to weather conditions and/or artificial noise. Therefore, an alternative computer vision system is required that can replace GPS in case of interference. For devices moving slowly, satellite navigation is inferior to video-based navigation. The same applies to angular positions.
This is why the Chinese are interested in this technology. They need drones to deliver food, mail and other small packages. Right now drones require dispatchers to control them because navigation failure or errors may result in an accident. But when drone delivery becomes common it will become uneconomical to hire too many dispatchers. There are several companies in the world using video-based navigation. Sightec (Israel) and Scientific Systems (US) are two good examples. Still, there are no mass commercial applications of the technology although there are some military uses. For example, Tomahawk missile (US) uses video-based navigation to move and hit targets.
There are four types of computer vision programs. The first type applies when no data is available about the flight area (no maps or satellite images) so position is determined against the point of departure and as you fly you create a map of the area. The second type applies when there is a map of altitudes and you know where you are. The moving camera provides photos (images) of the same area taken from various positions. Using a computer, you can receive a three-dimensional image of the area, compare it with the map and find your current location. This is how terrain-based navigation works. The third type is navigation that uses images taken from a satellite or another craft. You compare the image taken by a drone with these images and get your location. It is also possible to combine all three methods when there are no maps, but using a drone flying around you can make a map yourself and use it for further navigation.
Kupervasser needs some 2-2.5 years to complete the video-based navigation technology. The device, including a camera and a memory card, will cost about $100. One camera is enough for the positioning of the drone, another two are required to make stereoscopic images of close up objects. No expensive lidars or radars are required. The only case where radar could be useful is when photos are taken in low visibility (such as fog or smoke), but it is not required to measure a distance. But even in this case, radar can be replaced with a camera that can see through fog and smoke, or see infrared light. Ideally, video-based navigation is an alternative to the satellite navigation, but the former uses the inertial navigation system as well, which provides the drone with autonomous features.
Computer vision for a lawn mower
This project is aimed at creating a video navigator for controlling ground-based robots using drones. The technology focuses on robotic lawn mowers; about one million of them are produced every year. To use a garden robotic lawn mower, you pull a wire to indicate the border and prevent the device from crossing it or falling into the swimming pool. The lawn mower moves in a random pattern and bumps into the wire, which prompts it to stay and operate within the marked area. The other type of large lawn mowers, which moves along sides of roads and fields to mow off weed plants, have to be controlled remotely. These solutions are not very efficient.
A video-based navigator will control the lawn mower from a high pole or an attached drone with a video camera installed. The system shows an aerial view of the plot; in case borders have to be indicated it can be done in the photo image, with no wires required. The easily controlled video navigator monitors the robotic device so that it stays within the indicated area.
“This is a brand new market; it will be revolutionary if we succeed in introducing our video-based navigator,” Oleg Kupervasser said.
The market of video-based navigators for unmanned aerial vehicles totals $1 bn per year, while the robotic lawn mower market amounts to $100 mio per year. The project requires investments of RUR 10 mio ($151,000) for completing the work on the product. At the moment, finances for video-based navigation systems are provided by Israel’s Ariel University, which received a state grant of $100,000 for the purpose. This is a university where Kupervasser teaches mathematics and video navigation. A total of $225,000 has been provided to Kupervasser’s team for creating two types of navigators.
The Israeli researcher and businessman is not quite satisfied with financial cooperation with Skolkovo, with no grants provided by the foundation. Yet, he notes considerable advantages of the opportunity to receive microgrants for conferences, advertising support and educational programs.
By Natalia Kuznetsova