Ad block

Engineering autonomy for machines

Your smartphone beeps on a Sunday afternoon and you put down your drink, tell your smart home assistant to pause the movie streaming to your TV and make for the door. It’s the burger and fries that you had ordered earlier. Delivered right to your doorstep by an autonomous mobile robot (AMR).

As you repair to the sofa and your movie, you give nary a thought about the autonomous processes that helped in making your burger order come together. No, you don’t dwell on how the wheat that went into baking the burger buns was farmed using autonomous tractors and combine harvesters, packaged using robotic machinery, and delivered to warehouses where AMRs helped move the inventory for a multinational retail chain. For you, it is humdrum that the restaurant where you placed your order was manned by robots collaborating with humans to flip your burgers and make final preparations for delivery.

This is not science fiction. While we were captivated by the pomp and show of self-driving cars, engineers went ahead and turned the scene above into reality.

That is why market research firms are predicting a bright future for autonomous machines. Technavio, for example, expects the global AMR market to grow at a CAGR of 24 percent from 2018 to 2022. The same firm has taken a closer look at the agriculture sector as well and forecast the global market for autonomous farm equipment to grow by over $62 billion during the 2019-to-2023 period.

And such developments are also the reason that the Consumer Electronics Show (CES) 2019 will feature at least 19 conference sessions on autonomous machines or robotics.

Future Present

If it still seems incredible that we should be standing at the edge of a future filled with autonomous machines doing our bidding and assisting us in jobs too difficult or tedious for consistent outcomes, then let us examine the products available today that make it possible for you to enjoy that Sunday afternoon hamburger.

Autonomous Farm Equipment: While your beef may come from well-fed cows whose behavior — reactions to hunger and disease — was monitored by vision-based AI using systems like those by Irish startup Cainthus, the wheat for the buns may have been farmed using autonomous tractors made by companies like John Deere. That company has announced this year a new plug-in electric and fully autonomous tractor, which allows farmers to groom their fields while laying out and retracting a 1 km onboard extension cord. The corded power delivery means lower costs over the company’s older autonomous model that uses lithium-ion batteries.

Autonomous Mobile Robots: AMR use is being piloted in warehouses as well as “last-mile delivery.” For instance, San Jose-based Fetch Robotics has a range of AMRs that use a combination of light detection and ranging (LiDAR) and 3D cameras to navigate unpredictable warehouse environments, adjusting routes in real-time to avoid obstacles. The company’s robots bring not just hardware but a cloud-based control system that uses a simple graphics interface to make them easy to deploy. This ease of deployment opens up the Robot-as-a-Service (RaaS) business model for Fetch.

Meanwhile, San Francisco-based startup, Marble, and US-based on-demand delivery company, Postmates, are building autonomous delivery robots. Yet another company, Kiwi, already operates its AMRs in UC Berkley and has recently expanded operations to UCLA.

Autonomous Production: While some manufacturing equipment vendors are working on building self-maintaining machines that self-adapt processes to suit changing environments, others, like Kawada, are working on more flexible production robots to assist humans. Their Nextage robot, for instance, coexists with human operators and collaborates with conventional industrial robots.

To flip your burgers, however, a restaurant might use Miso Robotics’ AI-driven robots for the kitchen. The company’s kitchen assistant, “Flippy,” helps grill, fry, prep and plate your food orders. Fast-food chain Caliburger is already using Flippy in California and plans to take the robot to over 50 worldwide locations within a year.

Home Assistants: At home, even vacuum cleaners are gaining a greater degree of autonomy to pick up the crumbs you dropped on the carpet. The Roomba i7+ not only cleans your home on its own but also empties the dirt collected autonomously.

Brains for Autonomy

What’s common among all of these companies and their autonomous machines is the processing power enabled by advanced systems like those from NVIDIA. For instance, Kiwi is using the NVIDIA Jetson TX2 module, while Marble has products with the NVIDIA Jetson TX1.

Kawada, on the other hand, is developing autonomous industrial machines using NVIDIA’s most powerful module yet: the Jetson AGX Xavier. The module was built to meet the needs of AI at the edge and is like a workstation — it delivers 32 TOPS — that fits the palm of your hand. The module has a 512-core Volta GPU on board with Tensor Cores to accelerate large matrix operations for AI.

If you are planning to develop autonomous machines, NVIDIA’s Jetson AGX Xavier Developer Kit will get you going with a comprehensive set of tools, including NVIDIA JetPack and DeepStream SDKs, as well as CUDA, cuDNN and TensorRT software libraries.

If you have already started your project, the Intel Neural Compute Stick 2 (NCS2) will make it easy for you to train and deploy AI on the edge. The self-contained neural network on a thumb drive has an onboard Movidius Myriad X vision processing unit (VPU) and plugs into a USB 3.0 port to seamlessly convert and then deploy PC-trained models to a wide range of devices, such as drones, robots and smart home devices.

More Than Just Brains

Autonomous machines are more than just computers on wheels. They are complex machines that must sense their place in their environment as well as their interaction with it using a variety of sensors, depending on the intended application. For AMRs, image sensors, LiDARs and RADARs help with most of the navigation related information. All robots also need intelligent power and battery management systems.

This means that your autonomous machine project will likely involve a wide range of components and subsystems from an equally wide range of suppliers like Nvidia, Intel, Analog Devices, On Semiconductor, AMS and Basler. And you will need help from experts to not only know all your component options but get advice that is rooted in autonomous machine design experience.

Interested in learning more?

For more information on this topic, or to get in touch with engineering specialist who can help answer any questions you might have, head to

Ad block