March 25, 2023

Just like how they did for the exploding smartphone market over ten years in the past, personalized infotainment working methods and open-source software program look like sweeping the automotive business. The Android Automotive OS has been making headway in lots of market niches, beginning with full-electric automobiles like Polestar a number of years in the past. It’s solely a matter of time till the neighborhood and ecosystem mature sufficient to develop into a critical drive for enabling cellular improvement on one more entrance: the automobiles.

Whereas Android Auto (a reputation simply confused with the subject I will likely be going over right this moment) and Apple CarPlay have had a long-standing within the area, they got here with a number of caveats and restrictions. These largely pertain to the truth that many features-to-be would depend on low-level entry to the {hardware} of the automotive itself. This proved to be troublesome, with each options providing a restricted set of human-machine interplay capabilities, comparable to a heads-up show (the place out there) and radio. With that in thoughts, the use case for offering apps for the precise OS working the automotive was clearly wanted.

The neighborhood and documentation are nonetheless of their infancy and don’t but present a deep dive into Android Automotive OS. Furthermore, the educational curve stays steep, but it surely’s positively potential to piece collectively bits of data associated to improvement and deployment. On this article, I try to do exactly that, all whereas emphasizing the MacOS facet of issues.


As a normal precept, Android improvement can both be finished on an actual gadget or a corresponding emulator. Given the delicate nature of granting functions entry to the precise automotive {hardware}, the app has to go the entire 9 yards with Google Play Retailer eligibility. On prime of that, it has to adapt to one in all a number of classes, e.g. a media app to be allowed within the AAOS system. The excellent news is that there’s a chance for an app to combine and match classes.

Thus, distributors supporting the brand new ecosystem (as of now, amongst others, Volvo and Polestar) opted for making a customized automotive gadget emulator that carefully matches the specs of the infotainment methods contained inside their automobiles. Regrettably, Polestar and Volvo emulators include proprietary code, are primarily based on older Android releases, and don’t but help the ARM structure, which is of particular curiosity to builders working with ARM-based Macs.

Whereas official AAOS emulators can be found in Preview releases of Android Studio (from the Electrical Eel model onwards), usually the duty at hand requires personalized {hardware} and parameters. On this case, a customized Android model would have to be constructed from supply.

Constructing from supply

Constructing from supply code is a time-consuming enterprise that’s not formally supported outdoors 64-bit Linux platforms (whatever the goal structure). With that in thoughts, selecting a devoted AWS EC2 occasion or a naked steel server for constructing the ARM variations of the emulator appears to be one of the best total answer for Mac builders.

A requirement for unofficial builds on Mac gadgets appears to be having a disk partition with a case-sensitive file system and in any other case following some further steps. I selected a devoted construct system as a result of, for my part, it wasn’t well worth the hassle to arrange an extra partition (for which I didn’t actually have the disk capability).

The selection of the bottom Android launch is essentially depending on the goal gadget help, nevertheless, for ease of improvement, I might suggest selecting a current one, e.g., 12.1 (aka 12L or Sv2). Mileage could range with reference to really supported variations, as distributors have a tendency to make use of older and extra secure releases.

After getting their palms on a improvement machine, one ought to prepare the build environment and comply with instructions for building an AVD for Android Auto. The overall workflow for constructing ought to embrace:

  1. downloading the source code – this will take as much as an hour or two, even with respectable connection and department filtering,
  2. making use of required modifications to the supply, e.g., altering the default VHAL values or XML configuration,
  3. working the construct – once more, could take as much as a number of hours; the extra threads and reminiscence out there, the higher,
  4. packing up the artifacts,
  5. downloading the AVD package deal.

Leaving out the utilization specifics of the lunch and repo for now, let’s check out how we will make the default AAOS distribution match our wants slightly higher.

Tailoring a tool

VHAL (Car {Hardware} Abstraction Layer) is an interface that defines the properties for OEMs to ultimately implement. These properties could, for instance, embrace telemetry knowledge or maybe some information that could possibly be used to establish a selected automobile.

On this instance, we’re going so as to add a customized VIN entry to the VHAL. This may allow app builders to learn VIN data from a supposed automobile platform.

First off, let’s begin with downloading the precise supply code. As talked about above, Android 12.1 (Sv2) is the discharge we’re going to go together with. It helps model 32 of the API, which is greater than sufficient to get us began.

With the intention to get sources, run the next command, having put in the source control tools:

<p>> repo init -u -b android-12.1.0_r27 --partial-clone --clone-filter=blob:restrict=10M</p>

<p>> repo sync -c -j16</p>

Partial clone functionality and selection of a single department be sure that the obtain takes as little time as potential.

After downloading the supply, find the DefaultConfig.h file and add the next entry to kVehicleProperties:

.config =
             	.prop = toInt(VehicleProperty::INFO_VIN),
             	.entry = VehiclePropertyAccess::READ,
             	.changeMode = VehiclePropertyChangeMode::STATIC,
 .initialValue = .stringValue = "1GCARVIN123456789",

An summary of HAL properties might be discovered within the reference documentation.


Having modified the default HAL implementation, we’re now free to run the construct for an ARM goal. Run the next directions contained in the AAOS supply listing – utilizing a display is extremely really helpful if connecting by means of SSH:

. construct/

lunch sdk_car_arm64-userdebug

m -j16      	                # construct the requisite partitions

m emu_img_zip                   # pack emulator artifacts right into a downloadable .zip

Notice the sdk_car_arm64-userdebug goal wanted for emulation on ARM-powered Macs. A car_arm64-userdebug variant additionally exists. Make certain to not confuse the 2 – solely the previous has emulation capabilities! Attempt working lunch with out parameters to see a full record of targets.

The -jXX parameter specifies the variety of threads to make use of whereas constructing the Android. If the thread rely just isn’t offered, the construct system will try to optimize the variety of threads routinely. Endurance is suggested, as even with respectable {hardware} sources, the compilation is certain to take some time.

The ensuing emulator artifact must be out there within the out/ listing below sdk-repo-linux-system-images.[suffix].zip to be downloaded by way of scp or your file switch consumer of alternative.

Operating a customized emulator in Android Studio

Now that we now have our bespoke emulator picture constructed, there’s slightly trick concerned in making it out there for native improvement with out creating a complete distribution channel, as outlined within the handbook.

First, find the ~/Library/Android/sdk/system-images/android-32 folder and unzip your emulator archive there. The listing might be given an arbitrary title, however the total construction ought to comply with this format:

|_ [your name]
   |_ arm64-v8a
E.g., ~/Library/Android/sdk/system-images/android-32/custom_aaos/arm64-v8a.

Second, obtain the instance connected package deal.xml file and modify the gadget title to suit your wants. A package deal.xml is added after downloading and unpacking the emulator sources from the Web and must be recreated when unzipping domestically. After restarting the Android Studio, Machine Supervisor ought to have an choice to use your brand new ARM image with an Automotive AVD of your alternative.

After efficiently working the emulator, a newly created VIN property must be seen within the Vhal Properties of Automotive Information. Good one!

Whereas studying VHAL property values is out of the scope of this text, it must be simple sufficient with a few Automotive library calls, and Google created an example app that does the very factor.

Downloading the above instance (CarGearViewerKotlin) is extremely really helpful – in case you’re in a position to construct and run the app on the emulator, you’re all set!

Facilitating AAOS improvement on M1

One of many issues I stumbled upon in the course of the improvement atmosphere setup was that the Automotive library was not being detected by Android Studio, whereas the app nonetheless builds usually from CLI. This seems to be a identified difficulty, with no official patch but launched (as of October 2022). Nevertheless, a easy workaround to incorporate a .jar of the Android Automotive library seems to work.

In case of working into any issues, import the library from ~/Library/Android/sdk/platforms/android-32/non-obligatory/android.automotive.jar by copying it into libs/ listing within the venture root and add the next directive to your predominant construct.gradle file, if not current:

	implementation fileTree(embrace: ['*.jar'], dir: 'libs')

As soon as the venture is re-imported into the IDE, Android Studio ought to have the ability to decide up the Android Automotive library for import and autocomplete strategies.

The Actual Deal

Emulators are adequate for testing functions, however what about actual gadgets, comparable to branded infotainment facilities? As talked about earlier than, at the least two main distributors (Volvo and Polestar) provide the built-in Android Automotive expertise out-of-the-box of their automobiles. System photographs and implementation particulars, nevertheless, are proprietary and require enrollment into their respective developer partnership packages. Polestar provides a free AVD that emulates Polestar 2 habits, together with the display dimension, body and {hardware} controls – alas, at the moment solely out there for x86-64 platforms.

One of many alternate options value contemplating is the set up of Android Automotive on an actual gadget – be it a pill or perhaps a Raspberry Pi platform. Some modules will nonetheless require virtualization, however switching to a bodily gadget could possibly be a serious step within the route of higher {hardware} compatibility.

All of the above considerations increase the query – find out how to get the app to work on an actual AAOS inside a automotive? I haven’t discovered a conclusive reply to that query, at the least one which received’t contain third events holding the precise documentation sources for his or her gadgets. It is smart that some doorways will keep closed to the final programming viewers as a result of safety implications of making apps for automobiles. Nobody, in spite of everything, would need their automobile to be taken management of by a rogue occasion, would they?

Closing ideas

Programming for Android Automotive remains to be an adventurous endeavor. Regardless that the system has been round since 2017 (with APIs open to public in mid-2019), official documentation can nonetheless really feel considerably inaccessible to newcomers, and the developer neighborhood remains to be in its budding part. This requires one to piece collectively numerous bits of official guides and normal Stack Overflow information.

Backside line: AAOS remains to be behind the diploma of engagement that the common Android working system has been having fun with to this point. The longer term is wanting vivid, nevertheless, with distributors comparable to GM, Honda, BMW, and Ford keen to leap on the automotive improvement bandwagon in years to come back. If that’s the case, the ecosystem will inevitably develop – and so will the neighborhood and the help it supplies.