ADAS and driverless car integrated solutions from Fusion Project

Artificial Intelligence (AI) and Machine Learning (ML) are ubiquitous for Advanced Driver Assistance Systems (ADAS) and autonomous vehicles. This involves an increasing number of sensors that generate large amounts of data that need to be processed in real time. In addition to simulations, ML models are deployed based on billions of bytes of data recorded in the field, but more is needed to further refine and improve the models in current and future vehicles.

Leveraging live and test vehicles is one way to get information, which can then be used for AI training in the cloud. However, the 2TB of data today’s vehicles can generate per hour is simply too much, so reducing the amount of information sent to the cloud is critical.

Over-the-air (OTA) updates complete the cycle by providing these vehicles with updated models. This is not a real-time cycle, where changes are made automatically based on current inputs, but a new model is generated, then certified and downloaded to the vehicle.

Such a process involves many hardware and software components. Therefore, the Fusion Project (Figure 1) has been launched to make this cycle easier to implement and maintain. The project's initial target is intelligent vehicle lane change detection, but over time it will address various aspects of ADAS and autonomous vehicles.

The creators of the Fusion Project include Airbiquity, Cloudera, NXP, Teraki and Wind River.

The initial group of companies covers all bases in the cycle:

Airbiquity: OTA management software

Cloudera: Data lifecycle solutions and cloud services

NXP: Vehicle Processing Platform

Teraki: AI for Edge Data

Wind River: Intelligent Systems Platform Software

The integration of hardware and software in the loop is ongoing. Developers also need to work with companies individually - different solutions can be used in the design.

The current framework is based on data acquisition support on NXP hardware running Wind River OS and Teraki’s analysis and compression support, which helps reduce the amount of information sent to the cloud (Figure 2). Cloudera’s cloud provides support for acquiring data from the vehicle and running software for further analysis and ML training. Airbiquity’s secure OTA updates complement the entire process by providing a more efficient and accurate update model.

Data acquisition begins on NXP hardware running Wind River OS (1) with analytics and compression support from Teraki (2). The massively reduced data (3) is sent to Cloudera’s cloud (4). This is used to refine the ML model, and the updated version is sent back to the vehicle via Airbiquity’s over-the-air updates (5).

While the loop is easy to understand, getting all the parts to work together is not easy. These companies have integrated systems that provide an overall solution, but additional hardware and software must be incorporated. Likewise, developers must work with each company individually, but know that the individual parts work together and have been tested in the overall solution.

“Automakers face ongoing challenges implementing complex technologies, such as those necessary for advanced ADAS and the next phase of autonomous vehicle functionality,” said Phil Magney, founder and president of VSI Labs and formerly co-founder of Telematics Research Group. “There are many facets to the next-generation data management technology stack that enables the continuous improvement and deployment of AI machine learning models, so automakers need vehicle-to-cloud solutions that leverage key technologies across domains, like the one created by The Fusion Project.”

As the complexity of vehicles continues to grow, integrated systems have become an integral part of delivering solutions in the automotive space. Leveraging these integrations can significantly reduce time to market and overall costs.