Jamie Walker, courtesy of The Australian, 5 October 2016, where the title reads “Drones do it better as the technology leaps ahead”
Haven’t heard of Andrew Tridgell? Well, perhaps you should have. In the micro world of unmanned aerial vehicle buffs, he’s a bit of a legend. The man known as Tridge is the guy to talk to about the future of flying robots. He is lanky, laconic and there’s a touch of the geek about him, which is certainly not out of place in the backblocks of Dalby on Queensland’s Western Downs, where the air fills with cactus moths and the buzz of tiny aircraft engines.
The occasion is the UAV Challenge Outback Rescue, a world-class proving ground for technology that’s on the leading edge of where unmanned aerial vehicles are going in public use. The idea is to set a demanding task and get some of the top thinkers and practitioners in the converging fields of robotics, artificial intelligence and software development to fulfil it. The original mission, back in 2007 when the biennial UAV Challenge began, was to get a bottle of water to a point in a distant paddock occupied by a dummy nicknamed Outback Joe. This was supposed to replicate a real-life search and rescue operation for someone injured in the bush.
If that sounded basic enough, it turned out to be anything but easy. Maintaining a data link, putting robotic eyes on the target, keeping the aircraft aloft in tricky weather took more time and work than anyone had envisaged. Tridgell’s Canberra UAV team went close to meeting the challenge in 2012 but it took until 2014 for them to take home the $50,000 prize; dozens of other entrants failed.
This year’s event attracts teams from Poland, The Netherlands, Canada, Thailand and Australia, whittled down to 10 qualifiers from a field of 60. Outback Joe is back, along with Tridgell’s Canberra crew of computer scientists and software experts. Their 1.5m UAV is kitted out with eight vertical-lift motors. If you have cast an eye over a play drone in an electronics store chances are that someone involved in the competition wrote part of its code.The task was ratcheted up several degrees in difficulty: not only did the teams have to pinpoint the location of the dummy from a set of general co-ordinates at Springvale Farm, they had to land the UAV close by, pick up a mock blood sample and ferry it to base — a round trip of 23km to approximate a medical emergency.
An hour was allowed and if a second UAV was used as a communication relay for radio or a data link, that too had to return in the allotted time. This proved too much. Forward Robotics of Canada, the first starter, set the scene when one of its machines crashed and the second lost comms. The Thais from Kasetsart University had the motor of their chopper flame out.
Micro Air Vehicle lab TUDelft built an eye-catching aircraft that lifted off as a conventional helicopter, rotated in midair and flew like a biplane. Unfortunately for the Dutch, it clipped the top of a 20m gum tree and became pinned in the upper branches. Poland’s JetStream team sat glumly as the storm clouds rolled in, unable to launch their hand-tooled quad plane because of a guidance system glitch. A UAV crashed near the coffee truck at competition HQ, sending spectators scampering. Two others went down.
CanberraUAV looked like repeating its breakthrough 2014 win until the relay machine, a miniature chopper, lost power and soft-landed in a field. But Tridgell saw enough as he guided the primary aircraft in for a perfect landing 42m from the dummy — “It was a nice round number, a bit of a geeky computer science thing,” he says — to get excited.
A new algorithm had picked out Outback Joe, standing vertically, from the tight lookdown angle. The computer had done the work better than any person could. As programmed, it had searched for something that wasn’t supposed to be there in the landscape and fired off a compressed thumbnail image. “You would have had Buckley’s of seeing it,” Tridgell explains. “Look out the window of a plane, blink, and you miss it.” He believes this search capacity should be in the toolkit of every emergency service in the country. Drones are cheap, reliable and uber-efficient. “We should be able to have a lamington sale for the local SES and the proceeds should be able to buy a good UAV,” he continues. “It would save lives and save money.”
One problem is that drones have an image problem. In the public mind they are associated with lethal military strikes in one dusty corner or another of the Middle East, or with the perceived nuisance or privacy intrusion of somebody’s toy hovering overhead with a video camera whirring. Tridgell’s early claim to fame was his pivotal role in developing popular online file sharing program Samba. His interest in drone technology was cemented through his involvement in the ArduPilot platform, remote control software that by some counts has been downloaded by one million users worldwide. Tridgell insists that the technology be open-source, meaning it is free and freely available. The teams competing in the UAV Challenge all use the program, though some cling tenaciously to proprietary information.
Through the years the 49-year-old scientist has knocked back lucrative offers to work for military organisations. “I don’t assist them and I would prefer that they not use it,” he says of ArduPilot. “But you have to be realistic … if you design a knife, people are going to use it.” The Australian Defence Force announced earlier this year it would spend $3 billion on giant unmanned aircraft, with a wingspan matching a 737 passenger jet, to conduct ultra long-range maritime surveillance.
A member of the CanberraUAV team who works on drone systems for the ADF puts it this way: “There is always cross-pollination” between military and civil research. Philip Rowse, a systems engineer involved with ArduPilot through his tech company, says the problem can be when governments don’t pick up the system, as is the case with Kenya, which has rejected UAV surveillance of endangered elephant and rhinoceros populations. Rowse, who grew up in the East African country, isn’t entirely convinced by the explanation that the Kenyan authorities were concerned the technology would fall into the hands of Islamist terrorists.
As the UAVs were turning on the thrills and spills outside Dalby last Wednesday and Thursday, regulator the Civil Aviation Safety Authority scrapped licensing requirements for drones and remotely piloted aircraft weighing less than 2kg, prompting the pilots federation to warn of a heightened risk of a “catastrophic collision” with an airliner.
No one really knows how many drones are aloft in this country, as licensing generally applies to those in commercial use, not the lighter models people buy for fun. The new system means operators of sub-2kg drones now notify CASA and are registered but avoid costly licensing. The agency says there were previously about 700 certified operators in Australia; 630 notifications came in on the first two days of business, a CASA spokesman tells The Australian. Operating conditions still applied, and unless specifically authorised, drones could fly only during the day, in the line of sight of the operator and not within 30m of people, above 120m in height or within 5.5km of a radar-controlled airport.
The rule change is designed to make it easier for the likes of real estate agents and media outlets to use drones for aerial photography, and probably won’t bring the sky down for all the pilots’ hand-wringing. A well-informed participant in the UAV Challenge points out that testing has shown it is impossible for a collision with a 2kg drone to disable a modern airliner.
In addition to the search and rescue applications, he sees a world of opportunity for UAVs in civil use. Drones guided by ArduPilot are monitoring active volcanoes and forest fires in the US, and emergency services there have started to pick them up as well. Jonathan Roberts, a professor in robotics at the Queensland University of Technology who judged the UAV Challenge, says similar international competitions were pivotal to the development of driverless car prototypes. A Brisbane firm, RF Design, is showing the way in the UAV sector. It supplied low-power data modems to the teams in 2014 and has gone on to sell 10,000 sets worldwide.
A representative of the government’s Defence Science and Technology Group was on-site last week and big-name companies were represented among the sponsors, including Lockheed Martin, maker of the new F-35 stealth fighter that’s to equip the RAAF. Brisbane-based Boeing researcher Brendan Williams, who works on the company’s autonomous and unmanned aircraft systems, ran the safety side of the competition as delegate for CASA. Boeing has partnered with the Queensland government and Telstra in a pilot program for drones to service the gas and coal-rich Surat Basin.
Asked about the company’s international program to develop unmanned military aircraft, he chooses his words carefully. “There is always a connection between technology and its use in a military context,” he says. Underlining the sensitivity of the issue, he stresses that this is a personal view, not Boeing’s, and his involvement in the UAV Challenge has nothing to do with his day job.
Some believe that it’s only a matter of time before the autopilot gives way to the robo-pilot and the cockpit becomes sole domain of the computer. The military is already headed in that direction, with the next generation of warplanes tipped to be pilotless. But what about airliners? Would you entrust your life to Hal the machine, with no flesh-and-blood pilot on hand if things went awry? “Absolutely,” Williams says. “But we won’t get to that point as an experiment. We would get there with a level of rigour and safety that would support it, so when it happens it will probably seem quite unremarkable.”