Driverless cars are coming to our cities and streets, but first they have to find their way through some significant legal and policy challenges, Cyrus Shahabi writes.
The future is here, or at least, a short drive away: Google’s driverless car prototypes have proven feasible to release humans from the driving responsibility. Now, the next question is, can we release the humans from the decision-making responsibility as well?
For example, currently the human driver decides which route to take from a source to a destination. Could we instead let the car decide which route to take? This may seem to be a much simpler responsibility than the responsibility of driving the car; however, it has a much more global implication than one may think. In fact it opens up a huge opportunity. Let me elaborate.
Current navigation applications such as Google maps and Waze focus on optimizing for a single user – the driver – when crunching numbers to find the shortest or the fastest path. This is of course fine as long as everyone isn’t using the same navigation system or if they don’t follows the app’s recommendation. That is, if the drivers are humans.
But once we replace the drivers with machines, they will also listen and follow the recommendations. Hence, the cars and the navigation software will end up playing catch up, with all the cars going where the software tells them to go, which will then mean that route will no longer be the fastest. So where, you may ask, is the opportunity?
The opportunity is that if we assume we can control all the cars, then our navigation algorithm could optimize for network flow. In other words, the goal is to maximize flow in the road network of a city. This new perspective on path planning can potentially significantly improve traffic congestions in major cities. Imagine a system that knows where all the cars are, where they are going, when they want to be there and when they get themselves in accidents. It may even know how an accident may impact traffic and when it will clear up.
Such a god-view of the city with all the past and current data can predict the future and can control every car towards a better future, i.e., less congestion. In fact, the more a priori information the system has, the better it can work. If you tell the system that you’re planning to reach the airport tomorrow at 10am, the system with very high accuracy can tell you when to leave and then it will make sure your car gets to the airport when you need it. Connect this god-of-the-navigation to the social network and your driverless car may even pick up a couple of your buddies on the way to the airport as well.
In fact, the more “sensing” of the real world and the virtual-worlds, the more effective the navigation-god will become. If it receives the traffic-signals’ data, such as when they’re red and when they will turn green, and for how long, it can better optimize the flow. In case of disaster, the navigation-god will avert your car, keep the flow away from the scene of the incident and evacuate quickly the cars stuck in the disaster area. In sum, the global view of all the driver-less cars will do much better than bunch of human drivers optimizing selfishly. Seems like an argument against the capitalist market economy!
Of course there are several social and policy implications, not least of all the privacy issue. In this brave new world, the navigation-god will know where every car is. But before you get too tense, remember that Google navigation and Waze already have that information to some level.
Another social consideration is the dangers of giving up control full to a software system. What if the system is hacked? Or the underlying algorithm has an undiscovered bug? These are key challenges which will require careful consideration by policymakers, but programmers and software engineers have started including failsafe mechanisms for their software, the same way that other engineers do for other infrastructures such as roads and bridges. A related social issue is that in case of a problem, how the legal system and the insurance companies should react. Can the system be sued? Is it the driverless car that is making all the decisions, and what is the liability of the driver (or should I say the passenger)?
Finally, it will be interesting to see how generally slow-moving governments adapt to this rapidly-evolving technology. Would the city transportation agency be able to feed high-quality data to the system to ensure its proper operation? Can they take advantage of the system to design new roads or modify them? For example, based on the system feedback, can a city’s transportation agency react quickly to change a carpool lane to a toll lane? Or perhaps add new bus lanes or subway tracks to modify traveler behavior and encourage public transportation vs. private cars to improve the flow even further.
The next ten years will see this rapidly-moving sector take a sharp turn from the pages of prototypes to the streets and highways of our cities. It’s up to our policymakers to make sure that driverless cars isn’t an industry that stalls or one that’s left waiting for the red light to turn green.