Now that most of our cars are connected, the next step is to use contextual information to put the right information in front of the driver to ensure the best possible driving experience. So what do we mean about context? And what are we doing with it?
Let’s define context as “the set of circumstances or facts that surround a particular event, situation, etc.”I like to talk about context in the automotive space as the additional information available to in-car systems that can augment the interaction between the driver and the system.
This information can be internal or external to the system, but let’s focus on the external information. Where does it come from? Modern cars contain up to 100 sensors, ranging from engine performance to crash avoidance, and this number is growing. In fact, these developments are substantial enough that today there is a conference focused solely on in-car sensors and electronics. Take a second to think about the vast amount of information that these sensors make available, and the potential that information offers. People are quick to jump to the conclusion that this means cars are getting smarter, but we have to do something with this information to make it meaningful. In practice this is difficult, as these technologies are often implemented by different teams. We have been exploring how we can use this information in a meaningful way at Nuance.
We often look at how this information can influence the advancement of in-vehicle infotainment systems (IVIS). The topics of driver distraction and safety have been debated at length by automakers, technology providers, analysts, and other groups, and rightfully so. We believe that driver safety should be the topmost priority for anyone working on an IVIS. Part of that responsibility is to continually think of new ideas to enhance safety behind the wheel. Recently, we tested the concept of an IVIS that was “aware” of a crash avoidance sensor in a driving simulator. Our system adapted its behavior based on information coming from the sensors. If the driver was in the middle of an interaction and an impending collision was detected, the contextually-aware system halted the interaction to help the driver avoid a collision. Half of our participants used this system, while the other half used a system that continued as normal, oblivious to the impending collision.
Our participants were completing drives where they were following a lead car and using the IVIS to complete various tasks. For this test, drivers were given the following prompt before driving:
“I want you to imagine that you are going to meet your friend John at a theater in Palo Alto, but you aren’t sure of the exact location. You are expecting him to send you a message with the address, but you need to start driving now to make it there on time. When I tell you to, use the system to see if you have a new message. If there is a message, listen to it, but do not reply. Once you get the address from John, use the system to navigate to the theater.”
During the drive, drivers used our IVIS to check for text messages when prompted. The system notified them of a new message and began playing it after confirmation. During the readout of the theater address, the lead car slammed on its brakes, coming to a complete stop. For participants with the standard system, the message continued to play. For those with the contextually-aware system, the message played a brief alarm, said “Hold on a second,” and paused until the driver had recovered and began to drive again. Participants then had to use only the system to navigate to the address provided. If they did not remember the address, they had to use the system to get it again. This is an important consideration because one goal of intelligent automotive systems and assistants is limiting the number of interactions with the driver to further mitigate distraction.
We ended up with some encouraging results: tests showed that drivers who used the contextually-aware system had a lower rate of collision with the lead car than those that used the regular system. They also could better remember information and needed fewer interactions to complete the task. Of course, these are early-stage findings and our teams are working on improving this concept further.
This is just one idea of how we can use contextual information for automotive virtual personal assistants to mitigate distraction and to create a better user experience, and we aren’t stopping here. We are actively looking at ways we can apply various types of information and context to an IVIS to accomplish this.