In this first part of this two-part series, I discussed why the automotive industry, particularly the incumbent OEMs, is facing a big data challenge. This challenge is becoming extremely acute as a result of the increasing adoption of EAC vehicles combined with Mobility Services (EAC+MS) and the torrent of data that will be generated as a result of this adoption.
In this post, I present how the incumbent OEMs can address this challenge. To do so, automakers must:
- Think strategically and own the big data strategy. They must then drive the execution of this strategy instead of relying on their suppliers for partial solutions
- Revamp the vehicle’s computer system architecture to create a unified computing and big data architecture.
- Establish and enforce data ownership rights among the appropriate constituencies.
- Create a data-sharing culture.
Own the big data strategy
Newcomers to the automotive industry, e.g., Google, Tesla, Uber, but also startups like Faraday Future, Atieva, Renovo Motors and Divergent Technologies, have realized the importance of big data and associated analytics and have been creating detailed strategies as they stated to design their vehicles and associated mobility services. For example, they have been determining what data to collect to enable autonomous driving, or how frequently to update predictive models in associated with self-driving cars, what data to capture in order to understand driver preferences and vehicle performance. And more importantly, they have been determining what data will give them 360-degree view of the vehicle and its occupants. They have determined which data-driven applications their strategy will include, which will be developed internally, which externally, who’ll be custom, and which will be off the shelf. They have also established sizable data science organizations that are working closely with the vehicle design teams for the refinement and implementation of their big data strategies.
The big data challenge is unique to the incumbent automakers. Incumbent Tier 1 and Tier 2 suppliers have started developing such big data expertise over the years as more of the automotive innovation shifted to them.
The emergence and increasing adoption of EAC+MS provides the incumbent automakers with a unique opportunity to develop a big data strategy that integrates in-vehicle and backend hardware and software, as well as all of the associated services. In designing such a strategy, automakers should not only look at the strategies of their suppliers, and those of automotive newcomers, but they should also consider the big data strategies of leading companies in other industries such as telco, aerospace, manufacturing, logistics, and maybe even financial services, insurance and consumer packaged goods.
Create a unified computing and big data architecture
Over the years, more computing was introduced in every automobile component. We have now reached the point where an average vehicle may have tens or even hundreds of microprocessors. Because these components are provided by different suppliers, with the incumbent automaker acting more like a system integrator, these microprocessors are cobbled together without a central architecture. For the EAC vehicles, automakers must create a new computing architecture that is rationalized and provides the computing power and networking that the various components, including the thousands of sensors, these vehicles will need. This architecture must be unified with the big data architecture that the automaker must develop to serve their vehicles’ needs.
The design of these architectures should be driven by the parts of the EAC+MS value chain the automaker wants to own or control. They must address both the in-vehicle data, as well as data that resides outside the vehicle, in data centers managed by the OEM, and its partners. To define and develop these architectures, the automaker will need to partner with the right companies, particularly startups. By doing so, automakers will be able to ensure that their EAC vehicles are properly instrumented and they have access to all appropriate data. Startups like Renovo Motors have already been working with some automakers and their Tier 1 and Tier 2 suppliers to create initial versions of such unified architectures.
Such an architecture must include services for:
- Collecting all pertinent data, whether it comes from a component, the passengers, or the transportation infrastructure, and whether it relates to the operation of the vehicle or any other aspect of the transportation experience. As it happens with every IoT application, not all of the data that is generated would need to be collected and managed. For example, autonomous driving requires the capture of far more data than the personalization of in-car entertainment choices to the driver’s tastes. Automakers must also learn to create data networks from the users of their vehicles. In addition to bringing them closer to the users of their vehicles, such networks will enable automakers to constantly collect data that can be used in applications such as real-time traffic alerts, real-time public transportation alerts, e.g., Waze, MooveIt.
- Managing the collected data. To date automakers have not paid much attention to data management. Data storage though constantly dropping in price was considered a luxury for most but the high-end vehicles. In most recent models, in-vehicle data management is being used for some of the data generated by the vehicle’s embedded systems, mapping data, contacts data, and a certain amount of entertainment content. EAC vehicles have significantly different data management requirements. More data and information derived from data, such as predictive models, will need to be kept on board to aid with the autonomous driving (navigation data as well as vehicle health data). Periodically, data will need to be uploaded from the EAC vehicle to the cloud so that it can be used for updating the various predictive models employed onboard. The data and models used for the personalization of the vehicle to the passenger’s preferences can be available over the cloud. Automakers will need to determine what data will need to be stored onboard and what data will need to be available from the cloud over a broadband connection. They will also need to determine what data will be analyzed by them and which by their partners. In other words, as data moves out of the car, it may not be going to a single cloud. The car’s owner and most importantly the passengers will need to know where the data ends up and where it is coming from.
- Integrating the collected data. Because of the quantity and variety of low-level data (multiple sensors, such a radars, cameras and lasers, to gather information about the surrounding environment) that will be captured by EAC vehicles and the type of higher-level information that is appropriate for the predictive models being used, it will be necessary for significant amount of data fusion and integration to be done on board. Such fusion and integration will require specialized software tools that one today finds in commercial and military jets, as well as enterprise applications.
- Communicating data. The discussions about connected cars typically use examples with entertainment and productivity applications, i.e., passengers accessing streaming music, video, email, etc. As we have seen through the use of our smartphones, the bandwidth needed to achieve acceptable performance of such applications is available through cellular networks. However, the bandwidth offered by such networks in order to transmit the volumes of data captured and needed by EAC vehicles is inadequate. To address this problem, wireless carriers, like Verizon, are starting to experiment with hybrid transmission methods that utilize both cellular and wifi networks. Automakers must establish strategic partnerships with wireless carriers in this area.
- Extracting information from data. This is a very difficult issue and an area of active research. Some companies led by Google are starting to use increasingly complex deep learning systems in order to improve the navigation capabilities of autonomous cars. Machine learning is also used for personalization and situational awareness, e.g., personalized in-car entertainment content presented during daily commute, vs information and entertainment content presented during a vacation road trip to the Napa Valley. In addition to being able to use the right technologies to extract the right information for navigation and personalization, companies must also make certain that the models running onboard the vehicle are kept current, and are properly updated, i.e, as one model is updated it doesn’t create inappropriate interactions with the output of other models. This is not only a communication issue but a bigger model validation and quality assurance issue that must govern the entire model update process so that the right model are always in use.
- Securing data. Much has already been written (and here) about risk of hackers accessing the data of autonomous car. And we started to see the first acquisitions of cyber security companies that focus on automobiles. The public’s concern about data security has also been listed as one of the reasons that could delay the adoption of autonomous connected vehicles. To safeguard against data, in addition to cyber security solutions, automakers will need to look into data encryption solutions. Finally, automakers will need to worry about data loss and recovery. Today, when something happens to my smartphone and I lose data I can go to my computer, download a backup, and I’m back in business. The data I lose in such an experience depends on how frequently I’ve been backing up my phone and how much of my data I keep in the cloud. Automakers will need to develop a corresponding solution to this problem because it will happen since EAC vehicles are computers on wheels.
Establish and enforce data ownership rights
There is significant fragmentation in the automotive world regarding data ownership. Automakers are not eager to share the data they have been collecting. Service providers in the automotive value chain have taken a similar approach. As a result, while we may have the opportunity to start establishing rich data value chains, we are not doing it. As I’ve been arguing, EAC vehicles will capture a great variety of data, with significant amount of this data being personal. For example, the automaker, much like Uber does today, will be able to know where I was transported to every single day, what content I consumed, what services I used while being transported. Not unlike what has happened internet services, the participants in this value chain, including consumers (maybe even starting with the consumers), need to know their rights with regards to the data, and who owns each piece of data.
Create a data-sharing culture
It is impossible to think that a single entity, such as an automaker, will be able to collect all the data that is necessary and useful in various transportation solutions. We are already seeing the development of a very rich startup ecosystem working in the broader automotive value chain, in addition to companies like Google and Apple that are collecting data from the infotainment systems they offer to automakers. Most of these startups, e.g., MileIQ, Renovo, Automatic, to name but a few, generate and collect interesting big data and are developing expertise in big data analytics. Automakers will need to appreciate the existence of this budding ecosystem and determine how to best share data with the right partners. Otherwise, I fear that these companies will gain a competitive advantage and lock the automakers out. The experience of telcos with smartphones is not an example that should be repeated by automakers.
The adoption of EAC+MS is making the incumbent automotive industry, and particularly the OEMs, that they must be in the information and big data business. This is proving a big challenge for the industry because the big data business is very different from the manufacturing business they’ve been to date. To be successful in the information business, the incumbents, starting with the OEMs, must own the big data strategy, create comprehensive architectures from the ground up, establish and enforce data ownership rights for the consumers to feel comfortable, and create a data-sharing culture by working closely with their partners, many of which will be startups.
You can find Part 3 here.
© 2016 Evangelos Simoudis
Where do you see platforms and ecosystems playing into your analysis? What are the platforms? Do the current big auto manufacturers have a shot at creating a platform? Is anyone (like Google) releasing their web services and APIs to establish themselves as a platform? How could a company (like IBM with their Watson platform) play in analyzing the data?
Automakers will need to think about in-vehicle platforms and out-of-vehicle platforms. There are already companies that are developing in-vehicle platforms (both for collecting the very low-level sensor data), NVidia provides one such platform, as well as fusing the sensor data into more information-rich features. I feel that automakers will need to start partnering with startups that have been thinking about in-vehicle platforms. The existing big data platforms are more appropriate for out-of-vehicle processing, i.e., the data center. In this area, automakers may be able to adapt and utilize existing big data platforms that have been developed by smaller private companies and large vendors, like IBM. It would be a great idea if automakers can agree and define an API that provides richer data than today’s ODB to which companies, including startups, can write to. I’ve been arguing for such an API but I haven’t seen any movement in this direction yet. This is why in the piece I note the need for a culture of sharing.