Autonomous drive – cars and trucks that pilot themselves – is perhaps the most rapidly maturing areas of technology today. I know a senior IT executive at a major car company who, in the late 2000’s, served on a team dedicated to safety electronics like braking assist and lane monitoring for drowsiness detection. After moving to another department within the company, this executive returned to pay a visit to his old team just a couple years later and saw how the number of such point solutions had since multiplied and evolved into a unified web of analytics-driven capabilities that – together – made autonomous drive a reality.
Algorithms are behind the complex AI systems that now let cars – properly equipped with sensors, navigation and connectivity features – to drive and make countless traffic decisions all by themselves. In real-world settings, like Uber’s driverless taxi experiment in Pittsburgh, today’s autonomous vehicles are already using advanced analytics to manage autonomous drive’s three key functions of perception, prediction and motion planning.
Perception involves understanding what the car sees (Is that another car; a traffic light; a pedestrian?). Prediction helps the car understand what will likely happen next (Will that other car change lanes; is that traffic light about to change; will that person enter the crosswalk?). And motion planning involves decisions and task execution (Should we change lanes; should we stop; should we make a turn?) Capabilities are evolving and changing by the day.
The “we” in autonomous driving, however, is one thing that hasn’t yet changed: The ultimate responsibility today still lies with the driver. Each of those Uber taxis, for instance, has someone behind the wheel who monitors the car’s performance. Neither technology, nor the law, has yet reached the point where an autonomous ride can truly be “unsupervised” – the gold standard of autonomous driving that means people on board the vehicle are no longer expected to have responsibility for what the car does. That’s the next major frontier, and analytics will get us there soon.
Volvo, in fact, predicts it’ll be selling its first “unsupervised autonomous vehicles” to the public by 2021. The embodiment of international standards for full automation in driving, unsupervised autonomous drive is a big threshold that will be crossed only with high levels of confidence in the underlying analytics and systems that all work together in concert.
Sentient Cars…and Companies
It’s worth mentioning that this data-driven autonomy in vehicles is analogous to what happens on a macro level for an entire organization in the advanced stages of the Sentient Enterprise capability maturity model I created with the Kellogg School of Management’s Mohan Sawhney.
Fueled by advanced analytics, a “sentient” enterprise is proactive and can predict and adapt to changing circumstances – able to sense micro trends and make many decisions without human intervention. Such a company reaches these advanced levels of awareness and autonomous decisioning through seamless integration of various different analytic processes and architectures within the organization.
In a similar way, this “system of systems” interplay of algorithms and analytic processes is what powers a self-driving car to the point where unsupervised autonomous drive is now within sight. I already mentioned systems for perception, prediction and motion planning. All those capabilities will continue to improve, along with the systems governing connectivity, processing power – and the sheer electrical power – that will be necessary to support unsupervised autonomous vehicle operation.
A lot of work still needs to happen to make unsupervised autonomous drive a reality, but I don’t think we’ll have to wait long for it. As I mentioned at the outset, the pace of progress to date has taken even some industry veterans by surprise. If recent history is any guide, we’ll reach the next frontier of “unsupervised” sooner than you might expect – and we’ll have analytics to thank for it.