With self-driving cars generating so much buzz, it’s hard to believe that a self-piloting air taxi is, err, flying under the radar.
But not for long.
GTC attendees packed a conference room Tuesday to hear from Arne Stoschek, head of autonomous systems at Airbus A3 (pronounced “A-cubed”), the Silicon Valley-based advanced products and partnerships outpost of Airbus Group.
Stoschek is part of the team working on A3’s Vahana project, which aims to bring a self-piloted air taxi to the Bay Area’s skies.
The all-electric aircraft features rotating wings that allow propellers to be aligned vertically for takeoff and landing, and horizontally for flight. It’s designed to fly a single passenger or load of cargo over short distances.
The design makes airports unnecessary. The Vahana team envisions heliports atop buildings or parking structures. Each aircraft requires a landing area the size of two parking spaces.
A flight from San Francisco to San Jose, a trip that could take up to 90 minutes by train would last just 15 minutes. Even better, Stoschek said the cost of operating an air taxi is likely to be about $175 per flying hour, or a third that of operating modern helicopters. The total flight range will be about 110 kilometers, or just under 70 miles.
The company expects to fly a full-sized prototype and be in discussions about the path to legal certification this year. And Stoschek said a market-ready demo aircraft should be ready by 2020. He did not speculate about when the first commercial flights might occur.
The challenges still facing Vahana’s air taxi are similar to issues the self-driving car industry is grappling with. The most prominent: obstacle avoidance.
While there are far fewer objects in the air, both the air taxi and the objects it must avoid are moving much faster. So autonomous flying machines will need to detect objects from much longer distances.
And because all of the objects in question — primarily drones, birds and other aircraft — are often flying at low altitudes, the air taxi’s onboard sensors have to be able to detect them amid mountains, trees and buildings. Teaching the onboard system to detect every permutation of birds in flight is not an option, making them the biggest challenge.
“It’s a very tough detection problem, which makes it an ideal problem for machine learning” and, naturally, GPUs, Stoschek said. “We need to be both very accurate and very fast.”
The air taxi also must be able to navigate at very low speeds when near the ground, and avoid obstacles nearby.
Other technical needs to address include software and sensor optimization, even faster GPUs and reduced power consumption.
As for safety, Stoschek said that Airbus has a long history of manufacturing safe aircraft. Establishing fleets would be critical, he said, as they’ll be able to share information on all the lessons learned from encountering various flying conditions and obstacles.