Edge Computing: Vision and Challenges

Companies like Amazon, Microsoft, and Google have proven to us that we can trust them with our personal data. Now it’s time to reward that trust by giving them complete control over our computers, toasters, and cars. Allow me to introduce you to “edge” computing. Edge is a buzzword. Like “IoT” and “cloud” before it, edge means everything and nothing. But I’ve been watching some industry experts on YouTube, listening to some podcasts, and even, on occasion, reading articles on the topic. And I think I’ve come up with a useful definition and some possible applications for this buzzword technology.

What is edge computing?

In the beginning, there was One Big Computer. Then, in the Unix era, we learned how to connect to that computer using dumb (not a pejorative) terminals. Next we had personal computers, which was the first time regular people really owned the hardware that did the work. Right now, in 2018, we’re firmly in the cloud computing era. Many of us still own personal computers, but we mostly use them to access centralized services like Dropbox, Gmail, Office 365, and Slack. Additionally, devices like Amazon Echo, Google Chromecast, and the Apple TV are powered by content and intelligence that’s in the cloud — as opposed to the DVD box set of Little House on the Prairie or CD-ROM copy of Encarta you might’ve enjoyed in the personal computing era. 

Most of the new opportunities for the “cloud” lie at the “edge”

As centralized as this all sounds, the truly amazing thing about cloud computing is that a seriously large percentage of all companies in the world now rely on the infrastructure, hosting, machine learning, and compute power of a very select few cloud providers: Amazon, Microsoft, Google, and IBM. Amazon, the largest by far of these “public cloud” providers (as opposed to the “private clouds” that companies like Apple, Facebook, and Dropbox host themselves) had 47 percent of the market in 2017. The advent of edge computing as a buzzword you should perhaps pay attention to is the realization by these companies that there isn’t much growth left in the cloud space. Almost everything that can be centralized has been centralized. Most of the new opportunities for the “cloud” lie at the “edge.”

So, what is edge?

The word edge in this context means literal geographic distribution. Edge computing is computing that’s done at or near the source of the data, instead of relying on the cloud at one of a dozen data centers to do all the work. It doesn’t mean the cloud will disappear. It means the cloud is coming to you. That said, let’s get out of the word definition game and try to examine what people mean practically when they extoll edge computing.

Latency

One great driver for edge computing is the speed of light. If a Computer A needs to ask Computer B, half a globe away, before it can do anything, the user of Computer A perceives this delay as latency. The brief moments after you click a link before your web browser starts to actually show anything is in large part due to the speed of light. Multiplayer video games implement numerous elaborate techniques to mitigate true and perceived delay between you shooting at someone and you knowing, for certain, that you missed.

Edge computing has privacy benefits, but they aren’t guaranteed

 

Voice assistants typically need to resolve your requests in the cloud, and the roundtrip time can be very noticeable. Your Echo has to process your speech, send a compressed representation of it to the cloud, the cloud has to uncompress that representation and process it — which might involve pinging another API somewhere, maybe to figure out the weather, and adding more speed of light-bound delay — and then the cloud sends your Echo the answer, and finally you can learn that today you should expect a high of 85 and a low of 42, so definitely give up on dressing appropriately for the weather. So, a recent rumor that Amazon is working on its own AI chips for Alexa should come as no surprise. The more processing Amazon can do on your local Echo device, the less your Echo has to rely on the cloud. It means you get quicker replies, Amazon’s server costs are less expensive, and conceivably, if enough of the work is done locally you could end up with more privacy — if Amazon is feeling magnanimous.

Privacy and security

It might be weird to think of it this way, but the security and privacy features of an iPhone are well accepted as an example of edge computing. Simply by doing encryption and storing biometric information on the device, Apple offloads a ton of security concerns from the centralized cloud to its diasporic users’ devices. But the other reason this feels like edge computing to me, not personal computing, is because while the compute work is distributed, the definition of the compute work is managed centrally. You didn’t have to cobble together the hardware, software, and security best practices to keep your iPhone secure. You just paid $999 at the cellphone store and trained it to recognize your face. The management aspect of edge computing is hugely important for security. Think of how much pain and suffering consumers have experienced with poorly managed Internet of Things devices.

Bandwidth

Security isn’t the only way that edge computing will help solve the problems IoT introduced. The other hot example I see mentioned a lot by edge proponents is the bandwidth savings enabled by edge computing.

For instance, if you buy one security camera, you can probably stream all of its footage to the cloud. If you buy a dozen security cameras, you have a bandwidth problem. But if the cameras are smart enough to only save the “important” footage and discard the rest, your internet pipes are saved.

Almost any technology that’s applicable to the latency problem is applicable to the bandwidth problem. Running AI on a user’s device instead of all in the cloud seems to be a huge focus for Apple and Google right now.

Robot can pick up any object after inspecting it

 

Humans have long been masters of dexterity, a skill that can largely be credited to the help of our eyes. Robots, meanwhile, are still catching up. Certainly there's been some progress: for decades robots in controlled environments like assembly lines have been able to pick up the same object over and over again.

More recently, breakthroughs in computer vision have enabled robots to make basic distinctions between objects, but even then, they don't truly understand objects' shapes, so there's little they can do after a quick pick-up.
In a new paper, researchers from MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL), say that they've made a key development in this area of work: a system that lets robots inspect random objects, and visually understand them enough to accomplish specific tasks without ever having seen them before.
The system, dubbed "Dense Object Nets" (DON), looks at objects as collections of points that serve as "visual roadmaps" of sorts. This approach lets robots better understand and manipulate items, and, most importantly, allows them to even pick up a specific object among a clutter of similar objects -- a valuable skill for the kinds of machines that companies like Amazon and Walmart use in their warehouses.
For example, someone might use DON to get a robot to grab onto a specific spot on an object -- say, the tongue of a shoe. From that, it can look at a shoe it has never seen before, and successfully grab its tongue.
"Many approaches to manipulation can't identify specific parts of an object across the many orientations that object may encounter," says PhD student Lucas Manuelli, who wrote a new paper about the system with lead author and fellow PhD student Pete Florence, alongside MIT professor Russ Tedrake. "For example, existing algorithms would be unable to grasp a mug by its handle, especially if the mug could be in multiple orientations, like upright, or on its side."
The team views potential applications not just in manufacturing settings, but also in homes. Imagine giving the system an image of a tidy house, and letting it clean while you're at work, or using an image of dishes so that the system puts your plates away while you're on vacation.
What's also noteworthy is that none of the data was actually labeled by humans; rather, the system is "self-supervised," so it doesn't require any human annotations.
Making it easy to grasp
Two common approaches to robot grasping involve either task-specific learning, or creating a general grasping algorithm. These techniques both have obstacles: task-specific methods are difficult to generalize to other tasks, and general grasping doesn't get specific enough to deal with the nuances of particular tasks, like putting objects in specific spots.
The DON system, however, essentially creates a series of coordinates on a given object, which serve as a kind of "visual roadmap" of the objects, to give the robot a better understanding of what it needs to grasp, and where.
The team trained the system to look at objects as a series of points that make up a larger coordinate system. It can then map different points together to visualize an object's 3-D shape, similar to how panoramic photos are stitched together from multiple photos. After training, if a person specifies a point on a object, the robot can take a photo of that object, and identify and match points to be able to then pick up the object at that specified point.
This is different from systems like UC-Berkeley's DexNet, which can grasp many different items, but can't satisfy a specific request. Imagine an infant at 18-months old, who doesn't understand which toy you want it to play with but can still grab lots of items, versus a four-year old who can respond to "go grab your truck by the red end of it."
In one set of tests done on a soft caterpillar toy, a Kuka robotic arm powered by DON could grasp the toy's right ear from a range of different configurations. This showed that, among other things, the system has the ability to distinguish left from right on symmetrical objects.
When testing on a bin of different baseball hats, DON could pick out a specific target hat despite all of the hats having very similar designs -- and having never seen pictures of the hats in training data before.
"In factories robots often need complex part feeders to work reliably," says Manuelli. "But a system like this that can understand objects' orientations could just take a picture and be able to grasp and adjust the object accordingly."
In the future, the team hopes to improve the system to a place where it can perform specific tasks with a deeper understanding of the corresponding objects, like learning how to grasp an object and move it with the ultimate goal of say, cleaning a desk.
The team will present their paper on the system next month at the Conference on Robot Learning in Zürich, Switzerland.
VIDEO: https://www.youtube.com/watch?v=OplLXzxxmdA&feature=youtu.be

                                                                                                                     P.Revathi,M.Sc.,M.Phil.,
                                                                                                                     Assistant Professor,
                                                                                                                     Department of Computer 
                                                                                                                                             Science