Microsoft and the Automotive industry have a long history together, with Windows Embedded Automotive, Ford built their original SYNC platform, Toyota’s Entune and Fiat’s Blue & Me.
After originally launching in 1998, the latest Embedded Automotive release was back in 2010 and is due for the end of support March 1st next year.
With Windows Embedded Automotive done, Microsoft still believes they have plenty to offer the autonomous vehicle industry.
In a new whitepaper by Microsoft and Frost & Sullivan, business models, challenges, development, testing and solutions are all discussed.
As a cloud provider, Microsoft believes that can play a key role in the storage and analysis of the vast volume of data created by autonomous vehicles.
Connected, autonomous, shared, and electric (CASE) technology is expected to revolutionise the transportation industry, creating more dynamic, intelligent forms of transportation and new in-vehicle experiences.
Autonomous, or driverless vehicles have the potential to reduce travel time, ease road congestion, and deliver more reliable and safe mobility services. It also represents a major opportunity to improve the quality of life for the elderly, young, and disabled travellers that can access equitable mobility offerings.
Self-driving vehicles understand the environment around them and can reduce accidents, creating a path to actually attack road injuries and deaths.
Legacy automakers are struggling to adapt, and some OEMs are exploring new roles as service providers.
Some emerging business models enabled by AV are:
- Mobility Services: As it progresses towards full autonomy, the automotive industry is expected to shift from ownership to usership-based business models. Simultaneously, autonomy will help drive the convergence of the various fragmented mobility services that exist today
- Vehicle Services: As sensors and technologies become uniform across the industry, vehicle services will emerge as the differentiating factor. These services are becoming more dominant than automotive brands themselves and are expected to dictate the shape of future vehicles more than manufacturers.
- Unlocked In-Vehicle Experiences: With less time and cognition focused on driving, consumers will have the option to bring the office or home space into their mobility experiences, increasing productivity and comfort.
- Peripheral Services: Connected and AV data will be used to develop value-added services like usage-based insurance and predictive maintenance that can be offered to drivers, fleet owners, and mobility service operators.
- Logistics Services: New modes and mechanisms across various stages of the logistics chain will improve efficiency by optimizing time and cost of services. Some logistics providers are exploring autonomous delivery bots and drones to supplement traditional delivery services, while others have launched pilots to understand the feasibility of using shared mobility platforms to improve last-mile delivery.
- Long-Haul Delivery: Autonomous delivery represents a major improvement in long-haul efficiency, as vehicles can travel 24/7 with no human oversight or need to stop other than refuelling. As the industry continues to face a driver shortage, this space has the potential to yield high returns as shippers can instantly realize cost savings.
Advancements in autonomy also has an accelerating force on in adjacent segments such as tele-operations technology, and autonomous fleet management platforms. These new entrants often come from new tech start-ups are becoming more prominent in the automotive space.
Microsoft expects that automakers trying to keep pace with the competition in an autonomous future, must build toolchains that enable collaborate development, validate and manage AV solutions through deep learning, and provide full support through product life cycles.
Established automakers and new entrants alike are turning to technology partners that provide the most innovative and cost-efficient intelligent cloud and intelligent edge offerings to underpin their development toolchains. One of the best examples of this is Ford’s investment in EV startup Rivian for $500 Million to get access to their technology in an effort to catch up.
The overarching technology used to facilitate autonomy can be classified into 4 categories – perception, planning, actuation, and software. Perception allows the collection and processing of sensor and driving data, such as object recognition and localization.
Planning allows AI models to plan and predict paths, avoid obstacles, and control motion. Actuation controls braking, steering, and the human-machine interface (HMI).
Software is central to improving and deploying technology of perception, planning, and actuation. Software lays the foundation for end-to-end autonomous development along with each phase, from data ingestion and preparation to training and simulating models, and training and validating.
Given that dependency on software as a key competency, Microsoft are promoting the benefits of their Azure DevOps platform.
Microsoft goes on to set out some of the key challenges in building Autonomous Vehicles.
AV development requires a comprehensive toolchain, the various components of which seamlessly integrates and work together in harmony. End-to-End workflow management with built-in feedback loops and traceability for automotive certifications are two key challenges automakers are constantly faced with.
The other challenges include managing the massive test data generated by AVs, the massive compute needs required to train large sets of data, simulate real-world driving, validation such as software-in-the-loop and hardware-in-the-loop testing and accurate AI model training. AVs generate massive amounts of data during development. On-road trials are necessary to hone the AI powering AVs.
Depending on the level of autonomy, AVs generate 20-100 terabytes of data per day. Managing the sheer amount of data requires proper curation to filter and store data at scale and economically.
Naturally, Microsoft would love company’s to use their cloud services on Azure, expressing that on-site data centres will be hard-pressed to handle the storage and technology needs as the volume and permutations of ADAS, autonomous programs, associated simulations and validations increase.
Azure does offer a number of core services on Azure that could combine for a great back-end to AV development, offering the means to improve autonomous perception, planning, simulation, and deep learning.
When we break down what’s required to build an AV, you’ll need something like this:
Function | Product / Solution |
---|---|
Communciation | Teams / Office 365 |
Collaboration | SharePoint / GitHub |
Project Management | DevOps / Project Online |
Development | Visual Studio |
Prototyping (cameras, microphones, connectivity) | Azure IoT |
Testing | Azure Test Plans |
AI/ML/Computer Vision and Voice | Azure AI Platform / Machine Learning / Cognitive Services |
Storage | Azure Blobs / DataLakes / SQL Azure |
Customer Management | Dynamics 365 |
Web servers / Mobile app | Azure VMs / Mobile App Services |
While this is not an exhaustive list, it is fairly easy to prototype a model of Microsoft Services that would allow a company to have the necessary elements to build and deliver an Autonomous Vehicle.
I think it makes much more sense for Microsoft to play at this level of the AV race, rather than attempt to get back into a software platform development.
RIP Windows Embedded Automotive.
Of course, all these efforts assume you’re not going to try to build your own tech-stack of software and services like market leader Tesla. We know Tesla built their own autonomous hardware chip (HW3) that interprets inputs from the 8 cameras and outperforms Nvidia’s Drive platform, so yeah, good luck with that (a whole other specialisation).
You can read the who whitepaper here.