Each sensor has its purposes, strengths and weaknesses, and no one sensor can be solely relied upon to create a safe, fully autonomous vehicle. The use of software algorithms to verify the output integrity of multiple sensors and tie valid outputs together with GNSS and INS to create a high-definition environmental map is known as sensor fusion. This magic mix combines the sum of all parts to allow for the exponential increase in reliability that occurs when amalgamating multiple data outputs.
It's not only sensors dedicated to autonomy such as LiDAR, radar and cameras that are useful in sensor fusion. Most modern commercial and industrial road-based vehicles contain many other useful sensors, most coming from dynamic stability control systems or electronic stability control. Such sensors have existed for a long time but became mandatory in the U.S. in 2012 and the EU in 2014.
In a modern vehicle, you can expect to find speed sensors on every wheel, yaw rate sensors, steering angle sensors, transmission settings, throttle and brake sensors, all transmitting information to a high-speed data bus (Figure 54). Combining these sensor outputs alone with an onboard IMU can provide a positioning solution with redundant measurements for velocity, turn rate and vehicle direction.
While GNSS provides the only realtime absolute positioning solution, it is prone to loss. Unlike GNSS, IMU data is always available and can help provide vital relative positioning during a GNSS loss. The problem with INS is that it measures acceleration and rotation, not position, meaning the integration of acceleration and rotation must occur once to get velocity and again to get a position. Any acceleration or rotation rate measurement error will lead to an exponential growth in position errors. Additionally, like any microelectromechanical systems (MEMS) sensor, an IMU is prone to white noise and drift, introducing additional errors. Thankfully, sensor fusion can use the output from other sensors to constrain error growth and make a substantial difference in GNSS outage bridging.