The Technology Behind Self-Driving Cars

As recently as five years ago, the self-driving car seemed decades away from becoming a reality. Through the efforts of major automotive and technology firms, however, things are changing fast. Bringing their diverse expertise to the table, brands like Tesla and Google have gone a long way toward bringing autonomous cars to consumers.

But if you pay attention only to these big names, it’s easy to miss out on the whole story.

Much of the innovation that may make self-driving cars possible is taking place behind the scenes – the province of machine vision and imaging experts from an array of smaller companies. Arguably, vision technology is the most essential part of the driverless future.

The Nuts and Bolts of the Self-Driving Car

Machine vision is a main factor when it comes to making a car or truck autonomous. The vehicle needs to be able to respond to changing road conditions – becoming aware of its environment and taking action faster and more effectively than drivers.

This awareness comes from sensors integrated through networking technology, taking inspiration from the nearly ubiquitous presence of on-board GPS. There are many functions that work together, but they can be broadly categorized as follows:

Mapping & Localization – The First Step in Planning

Using laser rangefinders and cameras, the vehicle develops a map of its immediate surroundings. Rangefinders use laser beams to calculate the distance to nearby objects. Today’s advanced rangefinders see performance degradation beyond 100 meters, so on-board image filtering and aggregation are required to create a complete picture.

ROI Calculator

Discover the potential cost savings of robotic automation over a 20-year system life

This calculator compares your current manual labor costs against the total cost of owning and operating a robotic system over its 20-year lifespan.

EXPLORE TODAY

Obstacle Detection & Avoidance – Reducing Risk

In its internal map, the vehicle holds the current and predicted location of every nearby object – from static buildings and slow pedestrians to fast-moving vehicles. An on-board processor evaluates obstacle behavior against predetermined shape and motion profiles, allowing it to categorize obstacle types and respond appropriately.

Pathing – From Point A to Point B

No matter how sophisticated a map is, the map is not the territory – it has to be realized in action.  The most complex part of vehicle autonomy is finding and adjusting driving paths on the fly. That requires complex algorithms that account for road conditions and rules, chaining together short-range paths to reach a distant objective.

All these features come together to create what’s called a deliberative architecture. In effect, the data realized through machine vision feeds into true machine intelligence – AI that makes smart driving decisions in uncertain environments.

For autonomous vehicles to be viable, the entire deliberative process will have to take place nearly instantaneously no matter what happens out on the road. To make it possible, automotive engineers will need to work closely with machine vision experts now and long into the future.