As we move toward the era of autonomous vehicles, in theory, we will reach a time when we never again hear the bone-shattering, life-ruining crunch of metal that comes after the sound of squealing brakes. Self-driving cars will keep passengers safer – a large part of what makes them so appealing.
It is very easy to find encomiums of this future across the internet, including our own website. But what is rarely discussed are the people who are not in the cars – pedestrians and bike-riders, the people who are not plugged into the vast algorithms that will control driving and, unlike connected cars, will not be in constant communication with other vehicles. If the harmony between connected cars is a symphony, those outside the system are dissonant party-crashers. Yet, the system has to take them into account. Pedestrians must be kept safe and accounted for, an issue that is still being resolved.
Industry leaders are approaching the problem from different perspectives, all of which revolve around the relationship that drivers and those on foot currently have with each other. But uncertainty over the ultimate solution highlights the need for self-driving cars to be able to constantly improve via over-the-air (OTA) software updates.
Learning Human Behavior: Knowing When a Human Wants to Cross the Street
There is currently an understood pattern of recognition between a human driver and a pedestrian who is looking to cross the street, say, at a stop sign. Eye contact is made, one person waves the other on, and everything is figured out. Obviously, this cannot happen with driverless cars. To start, there will not be much need for stop signs, since cars will be able to communicate directly with each other, knowing when to stop without external signals. No stop signs or stop lights mean that traffic is, in theory, continual.
How then will a pedestrian cross the road? Should they just wait until there are no cars? What if it is a busy road? The current crosswalk and instructions to “wait for the signal” are less than ideal, as the whole point of driverless cars is to reduce or eliminate stopping time, making driving more efficient. Instead, technology must evolve to the point where waiting pedestrians can be recognized and factored into the system.
A “recognition” algorithm is already being developed by OEMs, which mimics instinctive human reactions to stimuli. Cars are being programmed to “see” if someone runs into a street or a deer bounds across the road and it is likely that they can eventually be programmed to recognize the behavior artificial intelligence patterns of someone wanting to cross the street. Artificial intelligence (AI) scientists outside the automotive industry are also at work on behavior recognition software, in which programs incorporate historical analysis to continuously learn. But this technology is still in its very early stages and today’s behavior recognition software is mostly used for analyzing patterns in markets.
To be effective for autonomous vehicles, algorithms must be able to understand things like jaywalking. While cars may operate on cold logic, humans rarely do, which means that we will essentially have to program irrationality into the algorithm. This kind of complex intelligence can take decades to fully develop.
Keeping Humans Alert
To quicken the process, engineers and futurists are exploring alternate ways of “looping in” pedestrians:
- Text messages: Some engineers think text messages represent a solution. You are waiting at an intersection, and you get a message that the nearest car’s algorithm has identified you and integrated you into its calculations. You are now free to cross. This could be inconvenient, however, since it would require you to have your phone with you at all times — although, in the era of earbud computers and other wearables, the message might be an instant ping or buzz. Still, requiring communication technology just to be able to cross a street would be an unfair burden on a large portion of the population.
- Lights on the car: It is not a big deal to have a light on top of each vehicle that signals “go ahead.” This would be a reassuringly simple way to get the message across while mimicking the “you go” gesture that human drivers now use. It would, however, require the car and pedestrian to be in fairly close proximity, requiring a noticeable slow-down in the smooth flow of autonomous traffic.
- Incorporating crossing signals into the infrastructure: Some futurists believe that lights in the road itself could signal when it is safe to cross. This system would not have to rely on a timer, like a modern-day crosswalk. A connected road would know when cars are coming and when they are not and could guide pedestrians across via a series of lights in the pavement itself.
In short, getting pedestrians to safely cross the street is both a minor problem and a major breakthrough. It is a small engineering task to add signal lights to a car, the breakthrough will be when cars learn how to interact not just with each other, but with the external environment — especially pedestrians. This is the core of AI – not just mimicking actions, but also understanding them, using “cognition” to be able to think ahead. Teaching an algorithm to apply critical thinking, to know that someone looking both ways is getting ready to cross, is the programming challenge of our day.
It will be a long and careful process, as automotive companies, tech companies, and regulators work together to figure out which of the available options would work the best. Progress will be incremental, and as the underlying software gets better, cars will have to be kept in sync through instant secure over-the-air software updates. All cars, and not just new models, must be on the same wavelength. This is the only way to make sure that cars are on the same page as humans – the non-programmed, non-algorithmic and most unpredictable component of the self-driving ecosystem.
As the auto industry is changed by technological and economic currents, OEMs and Tier-1 manufacturers will need to partner with technological specialists to thrive in the era of the software defined car. Movimento’s expertise is rooted in our background as an automotive company. This has allowed us to create the technological platform that underpins the future of the software driven and self-driven car. Connect with us today to learn more about how we can work together.