Decoding connected vehicle sensor data to deliver real value
Article By : Sumit Chauhan
As cars generate significantly more data every day, it's becoming a challenge to process, ingest and store all that sensor data efficiently in the car and to transfer data to the cloud. This article looks at converting that data into value for OEMs and the wider ecosystem.
The modern car has elevated from a mere source of transportation to a live connected vehicle and mobile data source that generates individualized information and scope for two-way communication. This data makes for one of the most precious values that can be derived from connected vehicles of today, forming the base for next-generation electric and fully autonomous vehicles.
A strong computing system undoubtedly becomes imperative to take advantage of this new-found glacier of vehicle and consumer data, and to learn what happens inside the car from the moment it is driven. Embedded vehicular systems kick start the data journey in connected vehicles, which when analyzed to its fullest give rise to new and convenient use applications, such as – optimized fleet management, data-driven usage-based insurance and smart battery solutions.
Embedded systems in automobiles
In purely technical terms, an embedded system refers to a computing device designed to access, collect, analyze (in -vehicle) and control data in electronic equipment, to solve a set of problems.
While embedded systems have been extremely popular in household appliances since the turn of the century, with the explosion in technology and microcontrollers, integration within vehicles has been fairly recent. While still in its adoption infancy, the technology has changed the outlook of the global auto industry, from design and manufacturing to safety and entertainment.
The data challenge
As cars generate significantly more data every day, it’s becoming a big challenge to process, ingest and store all that sensor data efficiently in the car and to transfer parts of that data to the cloud.
But not every piece of information in this data directly impacts user applications and thus serves little purpose if sent for storage on the cloud. Often the data relating to before or after an event doesn’t require cloud storage, and it is only the capturing of the exact event that needs cloud processing. Additionally, storing large amounts of data on the cloud comes with cost considerations and it makes sense for the OEMs to filter only useful and valuable data before it is transmitted to the cloud.
Why current solutions don’t work
The existing solution in the market is to use the low latency of 5G. Using AI and GPU acceleration on AWS Wavelength or Azure Edge Zone, OEMs can offload onboard vehicle processors to the cloud where possible. Traffic between 5G devices and content or application servers hosted in Wavelength zones do not traverse the internet, resulting in reduced variability and loss of content.
To ensure optimum accuracy and richness of datasets, and to maximize usability, sensors embedded within the vehicles are used to collect the data and transmit it wirelessly, between vehicles and a central cloud authority, in near real-time. Depending on the use cases that are increasingly becoming real-time oriented like roadside assistance, ADAS and active driver score and vehicle score reporting, the need for lower latency and throughput have become the need of the hour.
But while 5G solves this to a large extent, the cost incurred for the volume of this data being collected and transmitted to the cloud is still cost prohibitive. This makes it imperative that we have advanced embedded compute capability inside the car for edge processing to happen as efficiently as possible.
The utopian solution: vehicle to cloud communication
To increase the bandwidth efficiency and mitigate data latency issues, it’s better to do the critical data processing at the edge in the vehicle and only share event related information to the cloud. In-vehicle edge computing has become critical to ensure that connected vehicles can function at scale due to the applications and data being closer to the source, providing a quicker turnaround and drastically improving the system’s performance.
Technological advancements have made it possible for automotive embedded systems to communicate with sensors, within the vehicle as well as the cloud server, in an effective and efficient manner. Leveraging a distributed computing environment that optimizes data exchange as well as data storage, automotive IoT improves response times and saves bandwidth for a swift data experience. Integrating this architecture with a cloud-based platform further helps to create a robust, end-to-end communications system for cost-effective business decisions and efficient operations. Collectively, the edge cloud and embedded intelligence duo connects the edge devices (vehicle sensors) to the IT infrastructure to make way for a new range of user-centric applications based on real-world environments.
This has a wide range of applications across verticals where this data can be consumed and monetized for the OEMs. The most obvious use case is for aftermarket and vehicle maintenance where very effective algorithms can analyze the health of the vehicle in near real-time to suggest remedies for impending vehicle failures across vehicle assets like engine, oil, battery, tires and so on. Imagine being stalled on the road and an AAA roadside assistance comes up already handy with exactly what is wrong and the requisite vehicle part needed.
Additionally, insurance and extended warranties can benefit by providing active driver behavior analysis so that training modules can be drawn up specific to individual drivers based on actual driving behavior history and analysis. For fleets, the active monitoring of both the vehicle and driver scores can enable reduced TCO (total cost of ownership) for the fleet operators to reduce losses owing to pilferage, theft and negligence while again providing active driver training to the drivers.
Converting to sensor-based data into value
Once the cloud server receives ready access to data collected by the vehicle sensors, the next step in the process involves decoding this raw data into valuable information. Mobility intelligence achieved from unlocking the connected vehicle data, by analyzing it over past and current historic trends, has been successful in solving traffic flow problems in new ways.
With the implementation of a hybrid system, OEMs stand to save close to $5-$10 per month, per vehicle, in addition to the possibility for new use cases for monetization of the data that is still elusive for mostly cost and compute limitations as of today.
This demonstrates the huge value of such capability in the automotive industry. A hybrid solution with an effective mix of embedded synthetic sensors inside the vehicles, a distributed compute architecture and a powerful edge computing platform not just solves the data challenge but also the downsides of inefficient and slow communication. Hence these embedded systems in automobiles that assure high quality of data, in near real-time, can have far-reaching impact for automotive industries.
This article was originally published on embedded.
Sumit Chauhan is co-founder and chief operating officer of Cerebrum X, with more than 24 years of experience in automotive, IoT, telecoms and healthcare. Sumit has always played the leadership role that allowed him to manage a P&L of close to US $ 0.5B across various organizations, such as Aricent, Nokia and Harman, enriching their domestic as well as international business verticals. As co-founder of CerebrumX, he has applied his experience in the connected vehicle data domain to deliver the automotive industry with an AI-powered augmented deep learning platform (ADLP). Sumit is also passionate about mentoring and guiding the next generation of entrepreneurs.
