Sensor fusion and its role in safe autonomy

Autonomous vehicles require an enormous amount and variety of technology to work. More than that, they require safe technology with extremely low failure rates. 

Of all the required technology, the first point of entry is the sensor package installed on a vehicle. This package determines what a vehicle can sense about itself and its surroundings. Without this knowledge, an autonomous car simply cannot function. Add onto this the need to function safely, and it becomes clear that an array of complimentary and redundant sensors is required to guarantee the required level of reliability. 

The purpose of vehicle sensors can be divided into two categories: positioning (localisation) and perception. 

• Positioning is the act of establishing the position of the vehicle in a coordinate system, usually a global one. For autonomous vehicles, positioning also necessarily includes velocity and attitude 
• Localisation is a similar term often used interchangeably with positioning, though localisation refers to establishing the vehicles’ location within the local environment, rather than an absolute position in a coordinate frame 
• Perception is the act of detecting, identifying and tracking objects within the local environment relative to the vehicle

Whether for positioning, localisation or perception, the package of sensors must be combined to provide optimal results, in a process called sensor fusion.

What you will learn

This article explains what sensor fusion is and why it’s needed for positioning, and describes the types of sensors used to achieve the goal of safe autonomy, including ADAS, cameras, LiDAR, Radar, V2X and high-definition maps. Real-world testing results show how each type of sensor aids positioning in GNSS outage scenarios compared to SPAN GNSS+INS technology.

Read the full article by downloading Velocity 2023