Boston Dynamics rehabs drunken robot into parkour expert

Boston Dynamics has released a new video of its humanoid robot, Atlas, and it is scarily impressive how quickly the technology is advancing.

Two years ago the firm released a video of one of its employees testing out Atlas by taking it for a walk through snowy woods, knocking items out of its hands and knocking the robot off balance with a hockey stick. You can see the video at the bottom of the article, we decided the dubbed version was a bit more entertaining, though the latest video show remarkable progress by Boston Dynamics.

As you can see below, Atlas has gone from a staggering buffoon to a potential Olympian.

Many through the industry, us included, might be surprised. Robotics is an area which does not get much attention nowadays, perhaps due to the perception it is an area is not progressing in the same way as other next-gen tech. Of course, there are incredibly advanced machines in various different factories, but the idea of a humanoid machine we’ve gotten used to through Hollywood is laughable. Not anymore.

Although life for Boston Dynamics began as a spin-off from MIT, it was acquired by Google for an unknown price in 2013. Shortly after acquiring the robotics business, Google was reportedly in the market looking to offload the firm, struggling to see profitability or how it would fit into the wider Google universe. The rumours proved to be true, and Softbank purchased Boston Dynamics in 2017.

Atlas might be the product getting the attention today, but it is not the only robot worth keeping an eye on. Spot and SpotMini are four-legged versions, which move in a similarly remarkable way to Atlas, while Handle is a robot with both legs and wheels.

The incredible thing about Atlas right now is the progress which has been made in terms of balance. If you compare the two videos, the Boston Dynamics team seem to have mastered the art of stability and shifting weight in real-time. In the latest video, you can see Atlas even manages to identify the steps are not straight and shifts direction to safely land on the boxes. Such movement suggests the robots are not far away from real-world applications.

Stereo vision, range sensing and other sensors give Atlas the ability to manipulate objects in its environment and to travel on rough terrain, while it keeps its balance when jostled or pushed and can get up if it tips over. One of the immediate use cases could be delivery. There of course a few hurdles to adoption right now, the payload is only 11kg while battery life would have to be considerable, though these are minor challenges in comparison to what has been achieved so far. On the darker side, there will be several governments around the world looking at these robots with more militaristic ambitions.

There are of course scary implications for every new technology, just look at how the US Government wants to use AI to improve the accuracy of drone strikes, but that should not take away from the genuinely amazing progress here.

And as promised, here is Atlas in his early days:

Nvidia builds new AI platform to give robots better brains

Nvidia has announced the general availability of its Isaac platform designed to bring the futuristic world of robots for manufacturing, logistics, agriculture, construction and other industries.

The platform was launched at Computex 2018, includes hardware, software and a virtual-world robot simulator, as well as Jetson Xavier, which Nvidia claim’s is world’s first computer designed specifically for robotics.

“AI is the most powerful technology force of our time,” said CEO Jensen Huang. “Its first phase will enable new levels of software automation that boost productivity in many industries. Next, AI, in combination with sensors and actuators, will be the brain of a new generation of autonomous machines. Someday, there will be billions of intelligent machines in manufacturing, home delivery, warehouse logistics and much more.”

Looking specifically of Jetson Xavier, the box contains 9 billion transistors, delivering more than trillion operations per second, while using a third the energy of a lightbulb. Jetson Xavier has six kinds of high-performance processors, including a Volta Tensor Core GPU, an eight-core ARM64 CPU, dual NVDLA deep learning accelerators, an image processor, a vision processor and a video processor. This level of performance is critical due to the complexity of robotics with processes such as sensor processing, odometry, localization and mapping, vision and perception, and path planning.

On the Isaac Robotics Software side of things, Nvidia has billed the platform as a ‘toolbox’ for the simulation, training, verification and deployment of Jetson Xavier. The robotics software consists of Isaac SDK, APIs and tools to develop robotics algorithm software, Isaac IMX, the platforms Intelligent Machine Acceleration applications and Isaac Sim, a virtual simulation environment for training.

Nvidia will have a lot to live up for with these announcements. Aside from making big promises to a segment of artificial intelligence which has struggled to make progress, the team has stated the $1,299 box will have the same processing power as a $10,000 workstation.