Self-Driving Cars and “Learning” Computers: Securing the Future of Transportation
So often when we discuss the Internet of Things, automated machines, and integrating the Internet into every facet of our everyday lives, we focus on how people will adapt to these technologies. But what about the ways in which these technologies must adapt to us? In the case of self-driving cars, as this 60 Minutes special demonstrates, human beings are unpredictable creatures, and measures need to be taken for car computers and sensors to gauge dangers caused by human drivers and other potential dangers on the road.
Since self-driving, self-navigating cars operate through a number of sensors working on the inside and outside of the car, the software utilizing the data from the sensors must be able to “learn” human behaviour and other surroundings that may influence the safety of the car.
Mapping the World
A large part of Google's self-driving car software is an algorithm that maps terrain into virtual “tracks” for the car to navigate. These maps do not only plot out paths, roads, and highway systems — the map also acts as a full-scale analysis of every bit of geographic data it can gather. Every detail is plotted both visually and numerically; even the smallest details such as curb heights, streetlight locations, and minute road conditions are readied for the computers, sensors, and mechanisms of a self-driving car to use in real-time.
There is a problem with this, however. Plotting data on a scale this large is time-consuming and labour-intensive, especially with the data being as specific and readily usable as it needs to be. It is also, for the most part, regulated to urban and suburban spaces and highways. A logical next step for technology like this is using versatile IoT devices to transmit this data to car computers in real-time from wide distances.
Mapping Human Behaviour and Environmental Risks
While it is extremely laborious to map the United States' 4 million miles of road (let alone the rest of the world), and even more so to process that geographic data into usable information for the car's computers, it is only one step of the process in making self-driving cars as functional and intuitive as possible. What kind of variables can't be mapped? How will computers interact with variables such as animals and pedestrians on the road, natural disasters, or other things in the cars' immediate environment?
To return to the question posed at the beginning of this article, software and hardware need to be designed to better understand human behaviour and real-time environmental changes. Of the few accidents experienced by Google's self-driving cars, it has reportedly been the fault of the human driver involved rather than the self-driving car. Although these accidents are admittedly few and far between, the slightest risks with self-driving cars cannot only cause distrust and unease with a large part of a manufacturers' consumer base, but very real safety and security issues. The flawless integration of car, computer, and infrastructure is necessary to have safer roads on a wider, more global scale. Safer roads and more natural integration can only be accomplished by all of these factors working together and in constant exchange of data.
Self-driving cars will undoubtedly play a significant part in the future of transportation, and simply integrating the car to the Internet is not enough — both networks and automobiles must be more seamlessly integrated into the real world, and this involves designing software and algorithms that can allow these computers to learn about their environments.