Many of us try to juggle multiple tasks while driving—sending text messages, making an urgent phone call, sending or reading emails, etc. As traffic woes become more common, getting a few chores out of the way while on the go might not appear a bad idea, even though we’re well aware of how dangerous this can be. Even the smallest of distractions could be a matter of life and death on the road.
Some of the most common driving distractions are visual or auditory, fatigue, inattention, texting, eating, cognitive overload and drunkenness. These significantly increase stress levels while driving. In fact, 63% of Indian drivers report that congested traffic causes maximum stress, which includes worries on getting delayed for an important meeting or some other crucial appointment.
More examples of secondary tasks that cause distraction are adjusting cabin temperature, checking calendar, using speech commands, adjusting radio channels, updating Facebook/Twitter, listening to or composing a message, amongst others. An incoming call or message while performing a complex navigation manoeuvre is also a major stressor.
Today’s generation is inseparable from smartphones and expects to experience similar features in their vehicles too.
They want wireless connectivity, infotainment systems with high-resolution screens, web browsers, social media access and more. The cars that are on the roads today enable many of these functions. However, the problem is that the risk of meeting with an accident due to distractions caused by these high-end features is also significant.
So, there exists a major challenge for the automotive industry to provide consumers with the features they want, but in a manner that does not increase driver distraction and stress.
One way to mitigate this challenge is through ‘self-learning cars’. These are akin to intelligent computing devices that employ advanced machine learning techniques to learn the driver’s behaviour and preferences (without being explicit), predict driver needs based on this learning and overall driving context, and initiate proactive actions—much like a human personal assistant.
Such self-learning systems can assist the drivers in various ways—by easing common chores through an adaptive, contextual human machine interface (HMI), reducing distraction through proactive action on secondary needs of the driver, optimising speech and visual interfaces, providing contextual, crisp and just-in-time information, personalising the vehicle dashboard, etc. Over time, these systems become better through assimilation of deeper learnings about the driver.
In fact, there are some concrete ways self-learning cars can ease a driver’s secondary tasks.
* On identifying the driver when he/she enters, the vehicle’s infotainment system can predict the destination and route based on past learnings, and tell the driver the estimated travel time to the destination as well as inform about traffic incidents on the route, if any.
* Similarly, instead of just resuming playback from the last audio source (AM/FM, CD/DVD, USB/SD card) that the driver was listening to before switching off the ignition, the self-learning infotainment system can tune to the favourite news channel of the driver, depending on the time of the day, or act in a manner that aligns with the driver’s preference.
* To ease the hands-free phone dialling task, especially when the driver’s phone-book contains entries with duplicate names, the self-learning system can predict the driver’s calling pattern from past data to automatically resolve name ambiguity, and dial the correct number with minimal interaction with the driver.
* To reduce stress, the system could delay phone calls and other distractions when the driver is navigating a complex manoeuvre or is in a difficult traffic situation.
* Similarly, by extracting meeting information from the driver’s calendar, the self-learning system can trigger an ‘auto notification’ to attendees of the meeting, if it finds that the driver would get delayed for the meeting based on the estimated time to destination.
* By learning about the normal driving pattern of the driver, the system can detect when the pattern deviates from the normal, and alert the driver and co-passengers, if any, of the situation.
Some of the technical challenges that need to be overcome in realising self-learning systems are related to collection and sanitisation of data for training the learning system, seamless exposure of the capabilities of the learning system to the user as its capability improves over time, protecting the personal data of the user (behaviour and preference related), learning model deployment architecture (on-board, off-board or hybrid), etc.
Once these technical issues are sorted out, a self-learning car can truly play the role of an intelligent personal assistant-cum-smart virtual co-passenger, helping turn a stressful driving experience into a pleasant one.
By Debashis Mukherjee
The author is director, Harman India, a connected car technology company