Qualcomm’s Automated Driving Vision Gains Momentum with BMW’s iX3

Reaching the destination of fully automated driving vehicles feels much more like a marathon than a sprint. Some envisioned self-driving cars as early as 2020; however, it is much more complicated, with vehicles needing to learn and adapt to conditions in nearly the same way a human driver does.

In any case, a major shift is occurring between BMW and Qualcomm to help propel the transformation, with the new BMW iX3 leading the charge. This partnership exposes the Snapdragon Ride Pilot system, a platform that will help construct a "data flywheel" for the future of automated driving and could ultimately change the future of the occupation for the entire industry.

The Power Couple: BMW, Qualcomm, and the iX3

The introduction of the Snapdragon Ride Pilot in the iX3 was a true achievement, as it is not just another proprietary system housed in a particular brand, but rather, the launch of an open, scalable platform.

The two parties combined their strengths on the following:

Qualcomm's Hardware: The system is powered by Qualcomm's advanced Snapdragon Ride system-on-a-chip (SoC), offering the technological computing power to process complex driving scenarios.

BMW's Supercomputer: BMW provides the central processing unit as "Heart of Joy" supercomputer that looks after vehicle data and commands.

Collaborative Software: Together they produced a new stack of automated driving software. In the iX3, this is materialized by the Driving Assistance Professional package, and includes semi-autonomous features of driving with hands-free operation, on particular highways.

This partnership is the creation of a powerful system that can drive safely and smarter, because it can process a vast amount of vehicle sensor feedback in real-time for driving decisions.

What Exactly does the "Data Flywheel" Mean?

A flywheel is a physical piece of machinery that stores rotational energy, hence the more rotational speed it builds up, the more stability and energy it produces. 

This is what that looks like:

Data Gathering: The system, with the BMW iX3 fleet as a starting point, will aggregate sensor data from real-world driving. This sensor data includes cameral and radar sensor data on road conditions, traffic flow, the behavior of other vehicles, etc.

AI Learning: The data will be sent into a centralized, cloud-based artificial intelligence (AI) engine. The AI will review the data to improve functionality of Advanced Driver-Assist Systems (ADAS). It will learn regional driving behaviors, understand unique road configurations, or improve features like autonomous emergency braking.

Learning Globally: The learning can then be used to continually update and improve the software stack.

Sharing Knowledge: The updates can then be shared with each automaker using the same Snapdragon Ride Pilot software stack.

The outcome is collective intelligence. A car running in Germany could learn from a car operating in a scenario running in California.As additional manufacturers and cars hit the road, the flywheel can accelerate, gather more data, and increasingly increase everyone's innovation cycle.

Scalability Empowers the Domino Effect

One of the defining features of the Snapdragon Ride Pilot is its scalability. It's not intended solely for high-end luxury vehicles. The architecture might be similar to that of an entry-level vehicle with a single camera, as well as a premium vehicle like the iX3 with multiple cameras and radar systems. This democratizing characteristic helps make ADAS technology available to vehicles on the road.

Addressing Data Image Concerns

The notion of vehicles collecting and transmitting data undoubtedly raises data image concerns; however, they are engineered to lessen these concerns. Qualcomm states, earlier in this week's summit you heard, the focus of the Snapdragon Ride Pilot program is sensor data; not personal data.

Vehicle data, not driver data: Data collected is confined to data about the immediate environment around the vehicle, namely road conditions, object detection, detect sensor conditions, and not related to or connected to the driver's identity or driving habits.

Anonymized and aggregated: When data is utilized to train the AI models, the data are anonymous and used in an aggregated way.

OEM control: Even though driving performance data is shared to progressively enhance the pilot, the vehicle manufacturer retains the rights ownership for their proprietary, and their specific UX and UI data. For instance, social context and infotainment settings are brand-specific proprietary aspects.

Overall - Your vehicle is contributing to other cars along on the same platform without infringing on your data image.

The Journey Forward

The joint work between BMW and Qualcomm on the iX3 is more than a feature within a new vehicle. This approach seeks transformation away from the traditional paradigm of limited and unique ADAS technology development, into the collaborative and open paradigm.