With the White House Frontiers Conference bringing President Obama and major innovators to Pittsburgh, it only seemed appropriate to talk about one of the most recent innovations in pervasive computing happening in our backyard—autonomous vehicles. Uber’s announcement that Pittsburgh would be the site of its self-driving vehicle pilot in August got mixed reviews. The ride-sharing giant was praised, but many people were skeptical of the program’s success, voicing concerns about the safety of the technology.
The U.S. Department of Transportation (USDOT) recently issued a series of guidelines in response to the commercialization of automated vehicles. These guiding principles recommend manufacturers and developers follow industry best practices as they move towards more concrete safety regulations, a framework for testing vehicles, and a plan to establish consistent regulatory laws in the future.
With safety dominating the conversation, what should we expect as this technology develops, and how should we approach taming the complexities that come with automated vehicles?
First, Let’s Get Back to Basics
If we look to Trillions™, the book our founders wrote exploring the future of the pervasive computing, it suggests going back to the basics—building blocks to be exact. As manufacturers and developers build autonomous vehicles, the computing devices incorporated into their design are essentially a system made up of components or building blocks. We can decide to make these components interchangeable or not. Similar to Lego bricks, interchangeable building blocks are compatible (we like to say fungible), and can be swapped in and out to make it easier to build and maintain. We see fungible computing as an essential element to the future of pervasive computing. Interchangeable hardware invites competitors to design compatible “blocks” for other systems, which increases information exchange between devices and empowers people to use systems in creative new ways.
The drawback to fungibility is it’s expensive. Even competing developers must agree on standards for how components interact, which means testing is more difficult because there are more alternatives to test. Running multiple tests cost time and money, which can be off-putting—especially when trying to stay ahead of industry competitors.
Building an Ecology of Information Devices
When it comes to autonomous vehicles and other Internet of Things (IoT) devices, it’s important to design for a trillion-node network (the environment where everyday products are connected to each other by microprocessors) as opposed to a trillion-node internet (the environment where everyday products are connected to the internet). This means developing peer-to-peer devices, not peer-to-internet-to-peer devices, in order to support an ecology of information devices.
So how do you build this ecology? At MAYA the way we integrate devices into an information ecology is through an information architecture (IA). An IA is a blueprint for how information is used, flows, and gets interpreted by users—which is especially useful when designing an interface. In an information ecology, the idea is to keep the information exchanged between devices efficient and organized. A well-designed IA can provide the structure to support effective information exchange and easily allow other devices to enter the ecology through its design.
What Does This Mean for Self-Driving Vehicles?
Most vehicles already have some type of radar to sense objects in their path, but vehicle-to-vehicle communication would allow vehicles to share a lot more information like road position, speed, and steering direction. This increase in data exchange could ultimately improve road safety. MIT has been working with General Motors over the past few years to further develop car-to-car communication, and allow cars to enter the trillion-node network.
As we move forward with automated vehicles, it’s important to find a balance between fungible computing and maintaining the security and integrity of a system. Planes, for example, have overcome vulnerable systems by manufacturing each aircraft with three different kinds of computers that “vote” on how to control it. This factors out any inherent flaws that might be due to how each computer is built. Although this is not a cheap way to build airplanes, it does significantly decrease the risk of system failure.
In the future, self-driving vehicles will be on roads beyond Pittsburgh. It’s important we think about how to integrate them into the existing information ecology to ensure a safe transition to autonomous technologies.