Building the smart car of the future requires putting technology in the driver’s seat. The most fundamental feature of a connected car is network connectivity, and this usually involves a wireless local area network (LAN). This allows the car to share internet access and data with other devices, both inside and outside the vehicle.
According to McKinsey and Cisco, researchers estimate that nearly 250 million cars will be connected to the internet by 2020. Vehicles will be equipped with new capabilities including smart sensors, connectivity modules, and big-data enhanced geo-analytical capabilities. The report states that, by the end of next year, connected car services will account for approximately $40 billion annually.
With a connected car outfitted with modern technologies that provide a wide array of new features and functions, our driving experience becomes radically more efficient, productive, convenient, and enjoyable.
Building the smart car of the future
Connected car applications can be separated into two categories: single vehicle applications, and cooperative safety and efficiency applications. Single vehicle applications include concierge features such as apps that leverage personal calendar details and public traffic information to alert the driver when to leave for an appointment to arrive on time. As the name suggests, cooperative safety and efficiency applications provide additional safety features for the driver, such as forward collision warnings, lane change and blind spot warnings.
The connected vehicles of the future will revolutionise the way in which insurance premiums are calculated, and how underwriters assess the risk of vehicles. Additional data from connected cars will allow insurance underwriters to more accurately measure risk profiles, allowing them to better tailor insurance policies to individual drivers. Data drawn from connected cars will also allow insurers to detect insurance fraud easily by providing precise data on vehicle usage.
By gathering big data through connected cars it may be possible to uncover a wide range of previously unknown relationships between different factors. As more data is being gathered, companies will discover even more about consumer behaviour. Is there a link between the type of music people listen to and which drive-through restaurants they visit, for example? This type of information could have a huge impact on which radio stations companies choose to spend their advertising budgets on. Companies that make use of this driver data will need to do so responsibly.
Putting technology in the driver’s seat
There are four technology considerations that should be central to any connected car strategy:
Artificial Intelligence (AI):
According to an IHS report, in 2015 the install rate of AI-based systems in new vehicles was only 8%; this number is expected to increase to 109% by 2025. Over time, AI techniques will confidently predict the impact of different driving styles and other factors on various car parts, based on data from thousands of drivers.
Real-time Data Processing and Machine Learning:
An autonomous driving car is able to gain an understanding of the vehicle’s position and circumstances by combining multiple sensor outputs from devices including radar sonar, GPS, cameras, and LIDAR. In total, about 4 TB of data may be generated an hour with these processes and it must be processed onboard the vehicle. The data platform, therefore, needs to support true real-time data processing and decision making for critical functions like braking or accelerating.
Big data generated by connected vehicles has the potential to be extremely valuable, providing insights into driver behaviour and vehicle health. We also know that machine learning is a continuous process, where analytic models are repeatedly improved and redeployed over time. It is possible to make predictions in different ways within an application or microservice.
Hybrid Cloud Architectures:
Hybrid cloud architectures can be easily leveraged for machine learning infrastructure. Training can happen with large sets of historical data in the public cloud or in a central data lake within your own data centre. In fact, model inference can be done anywhere. In many scenarios, it makes sense to leverage the scale and elasticity of public clouds and spin up new large computing instances to train a neural network for a few days and then stop the instances. The pay-as-you-go pricing model is a perfect one, especially for deep learning scenarios.
Relevant summary information will be driven to a centralised fleet management application in a data centre or cloud, where it is aggregated across many vehicles in order to analyse fleet performance and to anticipate maintenance issues. Vehicle-to-vehicle functionality will further require that the cars communicate in a peer-to-peer network that supports omnidirectional data movement.
Automatic machine learning, or AutoML, allows you to automatically build different machine learning algorithms with various hyperparameter settings and features to choose and deploy the best analytic model. With AutoML, you can build analytic models without any knowledge of machine learning itself.
It is clear that the process of training and deploying machine learning models has precipitated new, iterative development processes that require massive volumes of training data and the close collaboration between various stakeholders such as data scientists, application developers, data engineers, and data governance specialists.
A final thought
While some people are still hesitant to trust AI, machine learning, and autonomous vehicles, it’s evident that these technologies are more effective than relying on humans. Autonomous vehicles aren’t perfect yet, but the trends suggest they will continue to improve.
Research shows that customers are willing to switch manufacturers just to be able to use mobile devices and connectivity. It is clear that the connected car is emerging as the next generation consumer platform. The applications that will be built on the back of the platform will, therefore, warrant significant new capabilities at the data platform level.
As computation and data processing becomes ubiquitous with an increasing proportion of processing taking place in the car itself, the primary challenges will be to manage the massive volume and velocity of data being generated from different sensors. Facilitating real-time processing of that data using emerging computational frameworks like machine learning and connecting the car seamlessly with the data centre will also be imperative.
Interested in hearing industry leaders discuss subjects like this? Attend the IoT Tech Expo World Series events with upcoming shows in Silicon Valley, London, and Amsterdam.