Automotive Sensor Data Crowdsourcing: Leveraging Automotive Megatrends to Transform Connected Car Services

The role of connectivity in the car is reaching an inflection point. In the past, connected car functions were confined to telematics and infotainment functions, keeping the driver safe, secure, informed, and entertained with services that were largely superficial, or which were intended to emulate typical digital experiences (from smartphones in particular) in a way that was safe on the move. In the future smart mobility context, connectivity will play a mission-critical role, enabling driverless operation, fleet management and remote diagnostics/predictive maintenance. Embedded connectivity will cease to be a “nice-to-have” feature to inform or entertain passengers and drivers, and will be a “must-have” technology that has a tangible impact on personal mobility. These new applications/paradigms include over-the-air software, firmware and map updates, prognostics and predictive maintenance, Vehicle-to-Everything (V2X) communications, and massive sensor crowdsourcing.

The first green shoots of these transformative and impactful connected car services are already visible in the market, with the crowdsourcing paradigm set to take off in the remaining years of this decade. Sensor data crowdsourcing sits at the center of the fastest growing trends that dominate the automotive market—embedded connectivity and active safety—and enables application developers and service providers to build their connected car services on real-world insights provided by connected and sensor-equipped cars on the road. The applications for the crowdsourcing paradigm in automotive vary from improving existing connected car services to enabling brand new uses that would otherwise be unachievable, and these include on-street parking guidance, traffic detection, user-generated maps for autonomous driving, and the training of deep neural networks for autonomous driving.

Sensor data crowdsourcing is built on a wide range of enabling technologies, including embedded cellular connectivity (for applications that require quasi real-time feedback), connectivity management, various sensors, location intelligence, robust cybersecurity, artificial intelligence techniques for descriptive and prescriptive analytics, and cloud platforms that can aggregate and process sensor data, as well as disseminate useful services built on these data. Therefore, crowdsourcing solutions and platforms of varying scope have been launched by mapping and location intelligence providers, sensor technology developers, Tier One suppliers, Original Equipment Manufacturers (OEMs) and aftermarket connected car startups. Indeed, the level of interest in crowdsourcing expands beyond the supporting technology ecosystem, and even beyond the entire automotive industry to include adjacent markets such as the smart home and smart cities, where automotive sensor payloads can also be leveraged to improve services. In the broader Internet of Things (IoT)/big data context, the future driverless car will be a unique agent/asset, a mobile combination of powerful sensing, edge processing, and robust connectivity. Potential nonautomotive applications include the use of camera sensors to provide mobile, neighborhood security, and surveillance services.

Just as critical as the enabling technologies are the market conditions that will usher in aggregation of sensor data on a large scale, especially openness, cooperation with competitors on data sharing standards (i.e., cooperation on service delivery, rather than on specifications), and a minimum critical mass of contributing vehicles. The critical mass required varies according to the application in question, but in each case the quality of service improves with an increase in the volume and variety of sensor datasets. This has driven an increase in “coopetition” between competitive players, particularly at the OEM level, which can be seen in three principal trends. The first and earliest is the consortium ownership of HERE, with three OEMs contributing sensor data sets from their connected vehicles to the Open Location Platform. The second is engagement with initiatives like SENSORIS, which are specifically aimed at “smoothing out” sensor aggregation between competing brands. Thirdly, there is the emergence of connected car data marketplaces such as Otonomo and CARuso, which specialize in aggregating, normalizing, and enriching datasets from OEMs and fostering engagement with service developers through new revenue sharing business models.

There are a number of key features and characteristics that will define a successful platform for automotive sensor data crowdsourcing. As mentioned above, some players in this space have limited ambitions in the crowdsourcing paradigm, perhaps only looking to leverage the experience of the community in pursuit of a single use case or application, whereas others, such as data market places, must look to satisfy all of the competencies discussed below, either in house or through strategic partnerships. On a fundamental level, these connected car data marketplace startups are middleware platforms, which live and die according to their ability to attract data suppliers (OEMs) and data consumers (application developers and service providers). Important characteristics to support this strategy include:

• Openness and Neutrality: Particularly when dealing with OEMs that are shaking off legacy conservative attitudes toward data sharing, a level playing field is important. Within this, there must be a framework for OEMs to withhold certain proprietary data, countered by revenue share incentives for greater engagement.
• Initial Investment: Although not a valid long-term strategy, buying datasets directly from OEMs with funding from investors can “get the ball rolling” in terms of attracting interest from application /service developers. In turn, building up a network of developers can make revenue sharing deals more enticing to OEMs and other data providers. One of the great virtues of the crowdsourcing paradigm is its self-reinforcing nature.
• Normalization and Enrichment: The most important competence is the ability to disentangle for differences in reporting formats and frequencies between OEMs. Furthermore, enriching the data payloads from OEMs with location intelligence and context from datasets in other verticals serves to attract interest from application/service developers as the quality of insight that they can get from engaging with the platform exceeds that of directly engaging with the OEM. This is central to any middleware play.
• Compliance: As consumer concerns over privacy grow, crowdsourcing platforms must be unimpeachable in their approach to privacy. As the vast majority of OEMs have a global footprint, they will be looking for a partner with the bandwidth to satisfy a myriad of patchwork regulations with respect to cybersecurity, anonymization, and data protection techniques.

For a broader discussion of the sensor data crowdsourcing paradigm and the transformative services it can enable in automotive, please see ABI Research Report (AN-2605).