發表評論-藤兒-萬維博客-萬維讀者網（電腦版）

最早出於中世紀傳說。據說「The Holy Grail」聖杯是耶穌基督在最後的晚餐上飲葡萄酒的用具。後來，從亞瑪利太來的約瑟用此杯接被釘在十字架上的耶穌傷口裡流出來的血，因此為基地教徒所珍愛。

以後，「The Holy Grail」聖杯幾經轉手，最後竟下落不明了。遂引出許多以尋找「The Holy Grail」聖杯為題材的故事。於是，「The Holy Grail」聖杯現在常被用來比作“無處尋覓的稀世珍寶；努力想得到卻永遠無法得到的東西。”

......

想一想，現在你也在尋找「The Holy Grail」聖杯嗎？

------------------------------------------------

來源：Vancouver Sun 2016-08-06

Robotics firm seeks Holy Grail: To make drones 'see,' think

potentially fatal collision may be less than 10 seconds away.

Vancouver-based Iris Automation is working on one of the most vexing problems in artificial intelligence — to make commercial drone flight more reliable, effective and most importantly, safer.

Teaching drones to “see” the world as humans do and make wise decisions is nothing short of a Holy Grail in the robotics industry, but one that co-founder and CEO Alex Harmsen believes the company is edging toward.

Smarter drones would also be able to work more independently, without the constant attention of an expensive, paid human operator or operate reliably beyond the operator’s line of sight, even in complex environments.

“Humans live in a very visual world, so we really prefer systems that perceive the world as we do,” said Harmsen, who is residing temporarily in Mountain View, Calif., to participate in the Y-Combinator business accelerator program. “What we are building is computer vision, situational awareness and artificial intelligence that could guide any robotic system.”

The individual technologies required for such systems from sensors to computing platforms are maturing, but autonomous systems need to be as close to 100 per cent reliable as possible before industrial users will adopt the systems, he said.

The market for reliable drones in mining exploration, oil and gas and pipeline inspection is potentially vast. They are all multibillion-dollar industries, Harmsen noted. Goldman Sachs predicts the market for commercial drones will reach $100 billion in the next five years.

“There is a huge developing market for industrial drones and that’s where we are focusing our energy right now,” said Harmsen.

But there is justifiable nervousness inside the drone industry that eventually — perhaps inevitably — someone is going to die at the hands of an irresponsible or unskilled operator. Jail time and fines of up to $25,000 are meant to deter rule-breakers, according to Transport Canada, which investigated six drone incidents in 2013, 61 in 2014 and 97 in 2015.

On May 25, two pilots — one from WestJet Airlines Ltd., one from Air Canada — reported seeing a drone that was flying too close for comfort on their descent into Ottawa. In response, two CF-18 fighter jets under the direction of Norad were scrambled to investigate and track down the errant drone.

Two-and-a-half weeks later in Winnipeg, police were called to the city’s airport after a drone flew within 25 metres of a landing plane. In both cases, the drone and its operator were never found.

Iris raised $500,000 from investors in Vancouver and San Francisco in March to transition the company from development to commercialization.

Harmsen and co-founder James Howard founded the company in 2015 in a Kitsilano basement suite, just months after graduating from the University of British Columbia, where the pair kindled their interest in the field participating in drone-flying competitions.

Harmsen worked briefly as a programmer in NASA’s Jet Propulsion Laboratory creating 3-D vision systems to guide the Mars Rover, while Howard accepted an engineering internship in spacecraft engineering for the UK-based Spire Global satellite firm.

“We founded Iris because we could see this demand building and we felt that all the technology required to make this work was right there for us,” said Harmsen. “Everything is ready, except the situational awareness element that would make it safe and reliable. We knew we had the skills to build that.”

Pilot projects are getting underway with the company’s development partners — a delivery firm, a mining engineering firm and a pipeline inspection company — to see how the systems perform in the field.

“Drones are already semi-autonomous following GPS waypoints, so we are trying to get over that last hurdle to complete autonomy,” he said. “That means real-time situational awareness and decision-making.

“Iris is creating systems that detect and avoid stationary objects from mountains and trees to power lines and oil derricks, but also mobile threats such as planes and other drones. Autonomous drones could be a huge help in areas such as pipeline inspection or search and rescue operations, but we don’t quite trust them to go off and work by themselves.”

A GPS failure or an incorrect map coordinate can result in a crash if the drone is not aware of its surroundings, a potentially dangerous and expensive outcome outcome in an industrial setting.

Iris systems aim to create three-dimensional maps of the environment in real time, track anything moving in the environment and assess its threat potential.

“At that point we can return control of the drone to an operator to avoid the problem or just cut power to the engines if it’s really serious,” he said. “Better to drop the aircraft right out of the sky than to have a mid-air collision.”

Identifying moving objects and plotting their trajectory is one of the key challenges Iris hopes to solve. The system’s camera-based sensor range is roughly 500 metres, a distance that would take a crop-duster less than 10 seconds to cover.

— With a file from Postmedia