There are over 100 million Alexa-enabled devices in the world, and customers love using Amazon Alexa for music, information, utilities, and more at home. With Alexa Auto, customers can enjoy the convenient capabilities of Alexa in their vehicles, too.
In mid-2021, the Alexa team saw an opportunity to provide automakers with even more options to build a multi-simultaneous wake word experience. The team began working with Panasonic Automotive Systems to build an industry-first integration for in-vehicle infotainment (IVI) systems that would give customers simultaneous access to both Alexa and Siri CarPlay. Amazon subsequently developed new APIs and software capabilities that integrate the Amazon engine for "Alexa" and "Siri" wake word detections to make the Alexa Auto SDK a turnkey solution for original equipment manufacturers (OEMs) to implement both Alexa and Siri CarPlay in their vehicles.
Giving customers the freedom to choose
In keeping with Amazon's philosophy of providing customers with more choice, the Alexa team has been focused on giving customers the flexibility of interacting with multiple voice assistants on the same device. In 2019, Amazon launched the Voice Interoperability Initiative, a group of more than 90 companies working together to bring this flexibility to customers.
"We believe that customers should have the freedom to choose their preferred voice service for any task," says Robert Mars, principal solutions architect on the Alexa Voice Services business development team, and Voice Interoperability Initiative device hardware and software lead at Amazon.
Over the last year, Amazon worked with Panasonic Automotive Systems to build simultaneous support for both Alexa and Siri CarPlay that automakers can add to their vehicles quickly and with ease. "OEM choices were limited in obtaining high-quality Siri wake-word detection," says Robert. "Alexa Auto offers it as part of the Alexa Auto Software Development Kit."
As a result of this collaboration, in December 2022, Amazon released an update to the Alexa Auto Software Development Kit (Auto SDK 4.2) that contains new APIs and capabilities for automakers to support multiple voice assistants with ease. Using the Auto SDK, automakers are relieved of months of development time and effort to integrate Alexa and other voice assistants into their products.
"We try to maximize the number of already developed capabilities that an OEM would otherwise have to develop on their own," says Robert. "We've made the Alexa Auto SDK a more complete package with multi-voice assistant capabilities so that OEMs can get their products to market more quickly."
Supporting multiple voice assistants
Alexa Auto SDK comes with three features that have traditionally posed development hurdles for automakers implementing multiple voice assistants: wake-word detection, speech region detection, and voice assistant arbitration. With the updates to the Alexa Auto SDK, automakers don't have to work with third-party companies to implement these capabilities.
What's more, the built-in wake word detector for voice assistants can support multiple languages simultaneously. In the case of Alexa and Siri CarPlay, when the languages supported by the vehicle's IVI system include the language setting on the customer's iPhone, keyword detection mode is used to recognize the wake words. However, when the customer connects an iPhone that's set to use a language the car doesn't support, the IVI system enters a voice activity detection (VAD) mode to detect speech in any language.
In VAD mode, when the IVI system’s speech region detector detects spans of voice activity that last at least 300 milliseconds, enough to be deemed speech, it transfers the wake word and speech request to the customer’s iPhone to process the wake word. Upon verification, the iPhone sends the request to its cloud services to respond to the request. While the IVI system is in VAD mode, Alexa continues to operate in the background. As a result, customers can use Siri in the language set on their iPhone and Alexa in the languages supported by the IVI system.
“It’s all about making sure automakers are able to provide as much choice and flexibility as possible for their customers,” says Robert.
New APIs for Alexa Auto SDK’s new built-in voice assistant arbitrator empower the application developer with a means of coordinating conversations between the customer and voice assistants. The arbitrator is a software component that coordinates multiple voice assistants and determines whether another assistant may begin an interaction while one is already active. When a dialog with one voice assistant completes, the arbitrator allows for a new interaction to begin.
There’s another important use case accounted for in the updates to the Alexa Auto SDK. The customer may want to interact with a different voice assistant while a voice assistant is speaking as part of an ongoing conversation. However, this is easier said than done because the rules related to barge-in vary among voice assistants. The Alexa Auto SDK allows the developer to specify the rules for barge-in for each of the voice assistants. The arbitrator interprets the rules and then applies the appropriate rule universally across all of the voice assistants on the device.
“The arbitrator provides a consistent customer experience by applying a single set of rules across all connected voice assistants,” says Robert. “Automakers don’t have to develop their system for enforcing these rules, which provides significant savings in development time and resources.”
In addition, vehicles may include a push-to-talk button that a customer can use to immediately activate a voice assistant. The Alexa Auto SDK arbitrator also accepts rules for voice assistant buttons and applies a common rule universally to these buttons to provide a consistent customer experience. In the more common use case, if there’s already an interaction with an active voice assistant in progress when the customer presses the button for another voice assistant, the arbitrator instructs the active voice assistant to deactivate immediately and simultaneously activates the pressed button’s voice assistant. Because the arbitrator manages these buttons, activating (and deactivating) any voice assistant is as simple as flipping a switch. This feature is especially beneficial in vehicles as it empowers the driver to stay focused on the road when interacting with multiple voice assistants.
Providing seamless access to both Alexa and Siri on the go
The Alexa Auto SDK simplifies development for automakers looking to provide customers with simultaneous access to Alexa and Siri CarPlay by providing a great “out-of-the-box” solution. In addition to using Alexa for in-vehicle controls, entertainment, and more, customers can say “Hey Siri” to interact with their CarPlay connected iPhone. Coordination between voice assistants is handled automatically by the arbitrator, and customers can use their vehicle’s push-to-talk button to interact immediately with either voice assistant. Customers have the freedom to interact with their preferred voice assistant and choose different voice assistants for different tasks.
“Automakers that use Alexa Auto SDK can provide their customers with a high-quality, proven implementation that supports multiple voice assistant systems working together,” says Robert. “They can offer customers a natural and consistent manner for interacting with simultaneously available voice assistants.”