Google algorithm lets robots teach themselves to walk

https://www.engadget.com/2020/03/02/google-algorithm-robot-teach-learn-walk/

There’s no question that robots will play an increasingly central role in our lives in the future, but to get to a stage where they can be genuinely useful there are still a number of challenges to be overcome — including navigation without human intervention. Yes, we’re at a stage where algorithms will allow a robot to learn how to move around, but the process is convoluted and requires a lot of human input, either in picking up the robot when it falls over, or moving it back into its training space if it wanders off. But new research from Google could make this learning process a lot more straightforward.

By successfully tweaking existing algorithms, researchers from Google Robotics were able to get a four-legged robot to learn how to walk forwards and backwards and turn, all by itself and in a matter of a few hours. First of all, they did away with environment modelling. Typically, before a robot gets the opportunity to learn to walk, algorithms are tested in a virtual robot in a virtual environment. While this helps prevent damage to the actual robot, emulating things like gravel or soft surfaces is extremely time-consuming and convoluted.

So the researchers began training in the real world from the get-go, and because the real world provided natural environment variation, the robot could more quickly adapt to variants such as steps and uneven terrain. However, human intervention was still necessary, with researchers having to handle the robot hundreds of times during its training. So they set about solving this issue, and did so by restricting the robot’s territory and having it learn multiple maneuvers at once. If the robot made it to the edge of its territory while walking forward, it would recognize its position and start walking backwards instead, thereby learning a new skill while mitigating human intervention.

With this system, the robot was able to use trial and error to eventually learn how to autonomously navigate a number of different surfaces, ultimately removing the need for human involvement — a significant milestone in making robots more useful. However, the research is not without its limitations. The current setup uses an overhead motion capture system to allow the robot to identify its location — not something that could be replicated in any real-world robot applications. Nonetheless, the researchers hope to adapt the new algorithms to different types of robots, or even multiple robots in the same learning environment, thereby creating a body of knowledge and understanding that will help advance robotics in all fields.

Via: MIT Review

Source: Arxiv [PDF]

via Engadget http://www.engadget.com

March 2, 2020 at 08:25AM

Developers And Publishers Say Google Isn’t Offering Them Enough Money To Bring Their Games To Stadia

https://kotaku.com/developers-and-publishers-say-google-isnt-offering-them-1842007925

Google Stadia hasn’t been a huge success. One big reason is the lack of games. And it seems the reason Stadia’s library of software is so small is that Google isn’t willing to spend the money to get more games on the platform. 

As reported by Business Insider, many developers have explained that one of the biggest reasons indies have mostly stayed away from Stadia is the lack of financial incentive from Google. One executive at a publishing company described the amount of money that Google was offering to them as “so low” that it wasn’t even part of the conversation.

Another indie dev described how most platform owners, like Microsoft and Epic, offer upfront incentives to entice developers and publishers to create or release games on their stores. But with Google Stadia the incentive was, in the words of that dev, “…kind of non-existent.”

Currently, there are only 28 games available on the video game streaming service. And while Google is promising to add over 10o to the platform this year, this report from Business Insider paints a picture that devs are nervous about working with Stadia. Many mentioned to Insider that they were concerned about Stadia’s future and how committed Google was to running the service long-term.

Since it was first revealed, Google Stadia has had a hard time convincing gamers to give the streaming service a shot. A lack of games, issues with pre-orders, old versions of games being uploaded to the service and other problems have plagued Stadia since it launched back in November 2019.

While new and big games, like the upcoming Doom Eternal, are still planned to release on Stadia this year, it does seem like the entire thing is on shaky ground.

via Kotaku https://kotaku.com

March 1, 2020 at 11:48AM

Riot’s Next Game Is Valorant, A First-Person Shooter

https://kotaku.com/riots-next-game-is-valorant-a-first-person-shooter-1842019029

Previously known solely as the League of Legends studio, Riot showed us last year that they’re now working on all kinds of new stuff. The first of which is Valorant, a team-based shooter.

We got a brief look at it in 2019, but here’s a full round of the game in action, captured with Riot devs playing internally:

The game’s website says Valorant will be 5v5 and pair gunplay with each character having special, unique powers, along with a very technical emphasis on lag and accuracy. Less scientifically, and with little to back this up aside from art style and the feel of it, the video is making me thinkthe planning for this was basically to land somewhere between Overwatch x Counter-Strike, which really, doesn’t sound like the worst idea.

It’ll be out this Summer in some form, according to the below tweet from the game’s official account. Whether that’s a full release or just widely-playable beta, we’re not sure yet.

via Kotaku https://kotaku.com

March 2, 2020 at 03:32AM

John Deere taps tractor-hailing tech for farmers in Africa

https://www.autoblog.com/2020/03/01/john-deere-hello-tractor-africa-sharing/

NANYUKI, Kenya/JOHANNESBURG — It’s ride-hailing, farm style. Deere & Co. is teaming up with the “Uber of tractors” in Africa and betting on a future where farmers summon machines with the touch of a button.

The world’s leading farm equipment maker is outfitting its tractors with startup Hello Tractor’s technology, which allows farmers to hail the machines via an app, monitors the vehicles’ movements and transmits usage information such as fuel levels.

The aim is to help the U.S. company boost sales of it famous green and yellow John Deere tractors, a tough task in a continent with the world’s highest poverty rate and the least mechanized agricultural sector.

Deere is currently testing the technology — a small black box fitted beneath dashboards — on around 400 tractors in Ghana and Kenya. It told Reuters it plans to roll out the devices across Africa in the second half of this year, offering it to all contractors who buy its equipment on the continent.

Jacques Taylor, who heads John Deere’s sub-Saharan Africa business, said that the continent badly needs more machinery to develop its farming industry but most farmers don’t have the scale to justify a large investment.

“We would like to see that every farmer has access to mechanization,” he told Reuters. “The gap that we’ve identified is, how do we connect small farmers with tractor owners?”

Deere declined to comment on the investment costs for the rollout. The risks are clear; there is no certainty of any measure of success in Africa, which accounts for a tiny fraction of its global sales at present.

Held back by low incomes, tiny landholdings as well as a lack of bank financing, tractor numbers have long been stagnant on the continent, even as much of the developing world has experienced a boom in mechanization.

Deere thinks it can help on the financing front: it told Reuters it could pull data from the Hello Tractor platform that showed in precise detail how farmers were using its equipment. That information, it said, could be used by the farmers — who typically lack credit histories — to help secure bank loans.

This would mean they could buy more tractors.

OPPORTUNITY KNOCKS?

In central Kenya, a Deere tractor zig-zagged across a sun-drenched field, raking up dry grass and dropping bales of hay. The black box monitored its every move.

The tractor belongs to Agrimech Africa, a Nairobi-based agricultural services firm that has taken up the offer to have the devices installed on its Deere machinery.

“They do the technology. We do the management,” said Pascal Kaumbutho, who heads the company.

Agrimech, which is paid by farmers to work their land, hopes the new tech will help optimize its Deere tractors and connect them to new customers, allowing it to expand.

Kaumbutho, whose company manages a dozen tractors, envisions a future in which Agrimech runs a 1,000-strong fleet. “Right now, we’re reaching about 1,500 farmers,” he said. “Within the next two or three years, I’d like to reach 20,000.”

Such opportunities exist in markets across Africa, said Hello Tractor founder Jehiel Oliver, but companies like Deere have lacked the tools to develop them.

“Nigeria alone needs 750,000 (more) tractors to be on the global average,” he said. “Our technology is a market-maker for tractor manufacturers who want to sell into those markets.”

Deere’s annual revenue of about $40 billion is dominated by the Americas and Europe. It doesn’t break out numbers for Africa, but combined revenue from Africa, Asia, Australia, New Zealand and the Middle East was $3.9 billion last year.

FINANCE FRUSTRATION

Outside South Africa, the continent’s most developed economy, around 80% of African cropland is still cultivated by hand. Yields are half the global average. With its population set to double by 2050, increasing productivity is a necessity.

One of the biggest barriers to mechanization is finance; though agriculture accounts for around a quarter of Africa’s economic output and some 70% of jobs, banks often view farmers as high-risk because of the lack of credit histories.

“It’s one thing to go to a bank and say ‘You know. Hey, I work very hard.’ It’s another thing to be able to show it,” Kaumbutho said.

Deere said the data from the Hello Tractor platform shows how often equipment is in use, how much land it’s working, and whether it’s tilling, planting or harvesting. That information can be used to create financial statements, it added.

Tshepo Maeko, vice-president and head of agrisales at South African-based lender Absa, sees potential to unlock more lending in this kind of technology which gives banks a fuller picture.

“We will be able to see how big the risk is or how big the opportunity is,” he said.

Deere is working with Hello Tractor and the banks to format the data to create easily digestible automated reports. No loan decisions have yet been made based on the information.

But Antois van der Westhuizen, John Deere Financial’s managing director for sub-Saharan Africa, said that should be possible by the time the scheme is rolled out across Africa.

“The banking systems are trying to adapt,” he said. “It’s a journey for us to really get them to understand it.”

Writing by Joe Bavier; Editing by Alexandra Zavis and Pravin Char.

Related Video:

via Autoblog https://ift.tt/1afPJWx

March 1, 2020 at 08:35AM