Industry Voices: How IMUs Help Make Autonomous Vehicles Safe

Opinion piece by James Fennelly product manager, inertial measurement systems at Aceinna.   

Autonomous vehicles are becoming a large part of the world around us. Today many vehicles are equipped with advanced driver assistance systems (ADAS) like lane departure warnings, assisted parking, automatic breaking etc, and although impressive, they are considered an SAE autonomy level of just 1 or 2, which means that they still require driver engagement 100% of the time.

So, when are we going to see the technological leap to Level 5, or fully autonomous vehicles, requiring no driver interaction? The truth is we are still several years away from this technology becoming widely accepted and adopted into any type of vehicle. The main reason is safety.

The final step in achieving full autonomy is 100 % confidence that a vehicle will continue to operate seamlessly in any conditions and avoid significant injury to passengers, pedestrians or property. This would require integrating a technology that ensures the vehicle will operate safely when other perception sensors like LiDAR, radar, or cameras fail or there is an intermittent disruption in GNSS satellite signals owing to weather, terrain or environment.

This technology is the inertial measurement unit (IMU). This will continue to send data to the vehicle so it will maintain a safe course until it can come to a secure stop or other navigation systems reengage. By eliminating data interruption and increasing operational safety, the IMU will expedite the reality of Level 5, fully autonomous driving. Without IMUs to provide the safety cushion, autonomous vehicles will never be able to effectively work in our city streets and highways.

What is an IMU Sensor and how does it work?

Most IMU sensors are composed of two different sets of sensors –gyroscope sensors and accelerometer sensors. The gyroscope sensors measure angular rate of three orthogonal axes. Integrating the angular rate along the three axes over time will generate roll, pitch, and yaw, which is the attitude of an object. Similarly, the accelerometer sensors measure linear acceleration in three orthogonal axes. Integrating acceleration over time will provide velocity and integrating velocity over time will yield distance travelled. An IMU with gyroscopic and accelerometer sensors can provide measurement over 6 degrees of freedom (6-DOF).

Why do some IMUs also include a magnetometer?

An accelerometer can be used to successfully calculate roll and pitch values with respect to earth’s gravitational force, and correct gyroscope drift. However, it cannot be used to detect absolute heading (yaw) because the change of yaw is orthogonal to the gravity vector. A magnetometer measures the magnetic field strength in three dimensions. Using the earth’s magnetic field it can help to determine heading (i.e., yaw) as well as roll and pitch of the object. Integrating a magnetometer in the IMU can help with detection of the initial heading of an object and correct integration errors of the yaw gyroscope in the sensor fusion algorithm.

Bias instability is one of the most critical performance parameters of the gyroscope. It is a direct measure of how much the gyroscope drifts over time. Because the rate output of the gyroscope is being integrated to calculate change in angles (roll, pitch, and yaw), any error associated with drifts results in accumulated error in relative angles. Furthermore, these angular errors translate into position errors over time. For automotive applications a high performance IMU is a necessary component for the autonomous vehicle to achieve high accuracy positioning.

Why is it important that IMUs are Triple Redundant?

In a triple redundant IMU, three IMUs are used to construct a triple-redundant sensor architecture that provides additional levels of reliability and accuracy.  If some reason, if one or more sensors is not working accurately, the system is smart enough to recognize the defective info and will not use it. The defective sensor output or errant dataset can be ignored or de-rated in importance. This architecture ensures the reliability of the system and simultaneously improves the performance.

Even though IMU sensors do not get the same amount of attention and press coverage as other sensors, such as LiDAR, radar and cameras, in many ways they are THE critical safety sensor component required for the successful operation of the Level 4 and 5 autonomous vehicles that will be appearing on our streets within the next decade.