This video identifies the hot spot use cases where edge analytics can be a superior solution to data-driven insights than traditional approaches or even big data-driven analytics.
- [Instructor] A great way to understand the value proposition of Edge Analytics is to look at the autonomous vehicle. Most every automobile manufacturer is working on autonomous vehicles themselves, or autonomous technology systems. They're joined by technology companies such as Intel, Apple, and Google's sibling company, Waymo. Depending upon where you live, you may have already seen autonomous vehicles out and about being tested. One of the most important aspects of the autonomous vehicle is that these cars need to do a lot more than just simply follow a pre-programmed route from point A to point B using GPS technology.
These vehicles must coexist on the roadways with other vehicles, autonomous as well as those controlled by drivers. The precise location, direction, driving speed, and other aspects of these other vehicles are dynamic and can't be pre-programmed. To enable this necessary coexistence, a number of different technologies must be built into and deployed on autonomous vehicles. Location sensing technology, typically GPS that we're so familiar with these days from our cell phones and tablet computers is, of course, used, along with many different digital cameras, radar and lasers, all of which help the vehicle understand its surroundings, including what is going on with other vehicles.
Sensors that are so common on vehicles today also come into play to understand what is going on with the vehicle's tire pressure, fluid levels, brakes, and other important subsystems. Sensors are also used to understand the outside environment, such as weather and road conditions. Even communications back and forth with other vehicles, for example, one vehicle indicating to others around it that it intends to change lanes shortly, will become increasingly important. Building and deploying the autonomous vehicle is, of course, about much more than just equipping a standard car with all kinds of advanced technologies.
All of these cameras and sensors and radar subsystems produce voluminous amounts of data on and ongoing basis. All of these data must be fed into a battery of sophisticated algorithms and models that make sense of what is happening within the vehicle and all around the vehicle, and then using the results of those algorithms to make numerous decisions about the vehicle's speed, ability to change lanes or turn, or perhaps even take evasive action on very short notice because of a hazardous situation that has just been determined.
Finally, these decisions must actually happen. In other words, the vehicle must take specific actions based on the decisions made by the algorithms. Examples of autonomous vehicle actions include, for example, following a planned route, perhaps over long distances without driver intervention, reacting to other vehicles on the road, for example, being cut off by some other vehicle and thus having to suddenly brake hard, or perhaps to take evasive action if a sudden stop can't be done safely.
If road conditions are hazardous because of snow or ice, for example, the vehicle will likely lower its speed and perhaps shift itself into a different driving mode, such as on-demand four-wheel drive, if that's available. Beyond the planned route from where we begin to where we want to end up, we often come across construction zones, lane closures, traffic jams, and other highly dynamic situations that could cause an autonomous vehicle to decide it must change lanes or perhaps even take a different route.
These decisions also need to be acted upon by the vehicle. Remember also that sensors are constantly monitoring the state of the vehicle itself. Fluid levels, tire pressure, and so on. And if a problem is detected, some sort of reaction to a given type of mechanical issue may need to occur. All of these actions are driven by decisions that come out of these algorithms, but where should these decisions actually be made? If you look at a centralized data lake architecture, you might think that using high-speed streaming of data from autonomous vehicles into a big data engine is what must occur.
The data lake will be equipped with decision-making algorithms and models and feed those decisions back to the vehicles themselves, hundreds of thousands of vehicles, perhaps, for the actions to be taken. The challenge with that approach, though is that latency, delays, however slight, between the time data travels from the vehicles to the data lake, and then the time lag between when decisions are produced from the algorithms, but then need to be transmitted back to the vehicles for the actions to be taken.
With Edge Analytics, however, we can push the decision-making power, the analytical algorithms and models, to the vehicle itself. This puts the decision physically closer to where those incredibly large amounts of data are being produced from the cameras and the sensors and the vehicle radar, which then puts the decisions in closer physical proximity to where those decisions need to turn into actions. In other words, what the vehicle must actually do, based on what it's being told.
The end result, latency is reduced, which is critically important for autonomous vehicles to operate as quickly as possible. Autonomous vehicles, therefore, are more like what you see here. Each one is equipped with very powerful analytics and no need to transmit data or to wait for data back from some sort of centralized data lake that was built using big data technology. Now, the question should be asked, then. Does this mean that data lakes have no role in the world of autonomous vehicles? Well, that's not exactly true.
Data lakes, indeed, may still exist as part of the architecture, with data being transmitted from the autonomous vehicles into some sort of a centralized big data environment. While this might sound confusing or perhaps even sound like double-talk, this actually makes sense. The autonomous vehicles will be making operational or tactical decisions from Edge Analytics. The steering and the speed of the vehicle at all times, along with specific actions that are necessary to avoid a collision with some other vehicle that's acting dangerously, or perhaps an unpredictable road hazard of some type.
Even beyond road hazards, taking a different route because of a traffic jam or some other situation is determined by the analytics within the vehicle itself. At the same time, the data lake receives data from numerous vehicles and may analyze those data for strategic purposes. The effectiveness of safety systems can be studied by looking at data and the success or lack thereof of avoiding collisions. The algorithms and models themselves inside the vehicle could be second-guessed or retroactively analyzed to help bring about continuous improvement in the Edge Analytics onboard the vehicle.
Product planning can also be supported by analyzing data strategically. Basically, our traditional dichotomy between operational and strategic business intelligence or analytics is still at work here, and understanding the operational side provides a great example of the case for Edge Analytics in many situations.
- Define edge analytics.
- Compare big data analytics and edge analytics.
- Identify examples of edge analytics at work in nature.
- Describe the stages of edge analytics frameworks.
- Explore the EdgeX Foundry framework for edge analytics.
- Provide examples of digital video data refinement and enrichment.
- List classes of analytics typically found in an edge analytics environment.
- Identify the objectives of edge analytics when applied to manufacturing.