Precise positioning for autonomous vehicles

Inertial measurement units (IMUs) are key to delivering highly precise position sensing for fully autonomous vehicles at a mass market price.

ByJames Fennelly January 6, 2022

Courtesy: Aceinna Inc.

Autonomous vehicles are being used in an increasingly diverse range of real-world applications. However, all applications using autonomous vehicles have a common need. All must know a moving object’s absolute and relative position in 3D space and how it is changing in real time. Some applications need more precise positioning information than others.

Inertial measurement units (IMUs) are used in many applications where absolute and relative position in three dimensions must be continuously estimated. Advanced driver assistance systems (ADAS) is one example. Others include robotics, drones, precision agriculture, and construction equipment 3D machine guidance/control.

Making highly precise positioning available to a wide automotive market looked impossible because it was too expensive. Now, though, emerginginertial navigation systems (INS) technologiesare changing the perception.

Figure 1: Some applications require greater accuracy than others. Lane departure warnings in passenger vehicles depend on accuracy within 10 to 30 cm. Precision excavation, on the other hand, works best when the desired position of the bucket’s teeth veers off by no more than 5 cm. Courtesy: Aceinna Inc.

INS’ role in autonomous navigation

INS use an IMU and sensor data captured by the global navigation satellite system (GNSS) to deliver precise absolute position, velocity, time (PVT) and attitude. PVT and attitude data from the INS are provided to the autonomous driving sensor fusion engine. Next, this data is further combined / fused with other perception sensors like LiDAR, RADAR, ultrasound, infrared, cameras, HD map data and V2X data to complete vehicle localization and navigation. IMUs are a key part of the autonomous vehicle sensor suite for the absolute positioning required to safely navigate autonomous vehicles.

Figure 2: Open IMU can be used in many different industries and applications. Courtesy: Aceinna Inc.

IMU fundamentals

Scientists and engineers have been interested in inertia for a long time. Galileo needed to refute those who insisted the Earth was not moving. The not-moving camp claimed the Earth couldn’t be moving, because if it were, then landing on the same spot that one jumped from would be impossible. The Principles of Inertia and Superposition resulted from Galileo’s efforts to counter this argument. [1]

acc的IMU测量或计算eleration, linear velocity, rotational rate, relative position and attitude possible. The IMU uses accelerometers to measure linear acceleration in all three dimensions. IMU gyroscopes measure angular (rotational) rate about the x-, y-, and z-axes. These are necessary for calculating change in velocity, position and attitude of a moving vehicle.

Integrating linear acceleration over time is how the vehicle’s relative velocity is determined. Once relative velocity is known, it can be integrated over time to yield relative position.

Integrating the rotational rate determines change in attitude (pitch, roll and yaw [heading]).

Figure 3: INS combines data from an inertial measurement unit (IMU) and GNSS position data and provides it to the autonomous sensor fusion engine for fusing with other perception sensors and high-definition map information to complete the localization of the vehicle. Courtesy: Aceinna Inc.

Every gyroscope has some degree of bias instability. The lower the bias instability, the more accurate the gyroscope and the longer it can safely navigate a vehicle even when all other sensors are offline (see below). To meet the needs of fully autonomous vehicles traveling at speed, the IMU industry is aiming for gyroscope bias instability of 0.1 degree per hour or less.

An IMU measures forces applied to it, including the force of gravity. In a perfectly stationary application, that is, one with no vibration or changing velocity, one could accurately determine the attitude from acceleration measurement of the Earth’s gravity and simple geometry.

However, other linear accelerations are present in dynamic use cases, and acceleration alone cannot be used to estimate attitude. In these cases, gyroscopes are used to measure rotational rate. Rotational rate is integrated with respect to time to provide change in attitude. Advanced IMUs employ proprietary algorithms and filtering to use rotation rate data in tandem with acceleration data to accurately estimate attitude and eliminate or correct for gyro sensor bias drift.

Figure 4: The three dimensions of linear acceleration and three dimensions of rotational rate measured by an Inertial Measurement Unit (IMU) with six degrees of freedom (6-DOF). Courtesy: Aceinna Inc.

An independent data source impervious to environmental conditions

As part of a localization solution that includes GNSS/RTK [2], an IMU stands out becauseonlygravity and motion affect it. An IMU is immune to environmental conditions that cause satellite signals to drop and to those that blind perception sensors. Inside a tunnel or parking structure when satellite signals are lost, during heavy rain or a dust storm, when camera lenses are muddy,no matterthe environmental condition, the IMU continues to function.

When satellite and perception sensor information is missing in action, the quality of the IMU is a factor in how long the IMU can guide the vehicle using dead reckoning. An IMU able to dead reckon for a longer period can do so because of improved bias stability over time and temperature, low noise and accurate scale factors.

Just to underscore why even the briefest interruptions in a precision navigation flow to an autonomous vehicle matter, at 65 mph (105 kph) a car would travel almost 100 feet (30 meters) in a second. Fortunately, even during a gap of as long as 30 seconds, an IMU can, as a self-contained independent data source, provide updates to position, velocity, time, and attitude for dead reckoning so the vehicle can be controlled and brought to a safe stop.

In addition to serving as the sensor of last resort in situations where all the other sensors have failed, an IMU estimates the vehicle’s position between each update of the GNSS/GPS receiver. The challenge is at highway speeds, the car could easily move more than 10 feet between GNSS updates. The IMU fills in the gaps between GNSS updates to identify the vehicle’s position and change in trajectory, pinpointing a car’s position at the 2 to 4 cm level.

Figure 5: IMU determines the performance of INS. Courtesy: Aceinna Inc.

IMUs and sensor fusion

The INS performs calculations to come up with the moving vehicle’s position based on data from the GNSS/RTK receiver and the IMU. While good GNSS/RTK data is available, the INS fusion algorithm is continuously estimating the current gyro biases and removing integration and other errors. When GNSS/RTK becomes compromised or lost, the INS sensor fusion algorithm uses data from the IMU to update the PVT and attitude until the GNSS/RTK signal is reacquired.

Figure 6: Unlike perception sensors (LiDAR, RADAR, cameras, etc.) an IMU sensor is not affected by environmental conditions like snow, rain, smoke, dust, etc. Courtesy: Aceinna Inc.

Future of autonomous vehicles

It’s reaching the point where it’s when and not if fully autonomous driving on a mass market scale will arrive.

The availability of calibrated microelectromechanical system (MEMS) IMUs capable of dead reckoning for more than 60 seconds and maintaining better than 30 cm of accuracy is one reason it’s “when” rather than “if” fully autonomous driving on a commercial scale arrives. Another is this level of positioning precision is becoming available at a size and price point in line with market demands.

Fueling those positive size and cost trends are two important phenomena.

One, there are advanced INS sensor fusion algorithms around application optimized Kalman filtering. These algorithms enable fast and accurate gyroscope and accelerometer bias, scale factor, and misalignment estimation. The result is continual error correction and accurate position estimates.

Two, leading IMU manufacturers understand how to capitalize on the size and cost reductions of MEMS sensors and increase production yields for accelerometers and gyroscopes.

Thanks to these recent developments, the wait time for the benefits of fully autonomous driving is shrinking, and IMUs can take much of the credit.

Figure 7: Comparing data sets within the IMU to increase accuracy. Courtesy: Aceinna Inc.

James Fennelly, product manager – inertial measurement systems,Aceinna Inc.Edited by Chris Vavra, web content manager,Control Engineering, CFE Media and Technology,cvavra@cfemedia.com.