Lightweight Edge Computing Scheme for Vehicle-Road Collaboration

Status of Edge Computing Industry under New Infrastructure

The mainstream MEC (Multi-access Edge Computing) edge computing products now used in the intelligent transportation industry lack top-down design and planning, and the open market competition may evolve into disorderly or even vicious competition, with the bad money driving out the good money, affecting the sustainability of the industry chain’s development, and the hundreds of thousands of data center servers are often used as edge-side devices, and users are helpless to pay for the ineffective computing power. Users have no choice but to pay for invalid arithmetic power, and the main arithmetic power is subject to the control of foreign manufacturers (such as NVIDIA AMD, etc.).

Currently facing problems:

The quality of middleware is worrisome, through the search for ready-made third-party modules to quickly stack the system, the vendor itself for the third-party module itself is the lack of assessment of the robustness, security, once the third-party components appear major security vulnerabilities, the vendor itself lacks the ability to solve the problem, and the consequences are worrying.

Hardware solutions focus too much on cost and short-term goal achievement, and are indifferent to the medium- and long-term protection of the user’s investment to the maximum extent. Insufficient assessment of the future need for elastic computing resources, poor functionality and performance scalability, once the demand for upgrading, the need to redeploy and batch replacement of equipment.

The lack of a unified platform, as well as the lack of standardization constraints, resulting in disorderly and vicious competition, set up a large number of unrelated to the actual application requirements of the “control point”, on the one hand, the user’s real needs have not been the best to meet, and on the other hand, the user in the unneeded needs to pay excessive costs.

The compatibility between embedded device platforms is poor because of the differences in hardware architecture and operating system. If deployed on a large scale, the black box nature of the device, and the demand itself requires the device to have openness, resulting in end-users susceptible to kidnapping by vendors.

The underlying operations between vendors are not compatible with each other, hardware device authentication management, health management, fragmentation of the situation, a set of projects a device management platform or even multiple device management platform.

Solution:

In intelligent transportation scenarios, MEC edge computing gateway is deployed on the roadside, which is able to quickly process and analyze data after receiving comprehensive data such as video images, and at the same time, through a variety of algorithms, such as vehicle detection, pedestrian recognition, speed measurement, etc., on-site real-time research and judgment to make the results of the process, which meets the requirements of real-time traffic, and synchronously transmits the data to the center, and is able to real-time prediction of the future traffic conditions;

In enterprises, parks, public places and other scenarios, such as monitoring the fire, smoke, riots and other public safety and other conditions, edge computing can be on the scene according to the learning experience the first time to execute the command, without waiting for the center to deal with the analysis of the command, direct and effective in the early stages of the critical situation to make a warning.

But the reality is constantly reminding us that with the advent of AI, algorithmic learning costs are getting lower and lower, these applications above really need dozens of hundreds of Tops of arithmetic support? Who should pay for the arithmetic card that is in short supply and tends to be in short supply? Hardware carrier tens of thousands of hundreds of thousands of costs how to better landing applications?

Thanks to the team in the image, control and communication business on the immersion and combat experience, Yingli for a single point single scene launched the “lightweight micro edge gateway”, we believe that low-cost small arithmetic can make the arithmetic deep into the edge of the network of the end of the micro, using our program to let the application of the edge side of the advantages of the edge side of the full play.

The main features of the program:

Equipment vendors as users of the industry’s lightweight micro-edge computing module, which facilitates the improvement of the system’s standardization and security, and avoids disorderly repeated investment and vicious competition.

Core chips, core components (such as file systems, databases, tools and software) are localized, and the source code level is independently controllable.

Adoption of independent intellectual property rights operating system for industry-specific secondary packaging, as a unified system base for continuous iteration, to avoid fragmented development.

Adopting whitelist application market management to ensure the effectiveness of the overall system security policy, with industry users as the first principle.

Unified cloud platform management for device and application authentication, effectively resisting deception from the physical layer to the application layer, and facilitating scaled automated deployment and maintenance.

According to the industry application to customize the suitable computing power equipment (such as a single device to deal with only 5-10-way video), based on low-cost solutions to improve the computing power points, and truly realize the edge-side computing applications.