What is an IMU

Overview

The documented heritage of this site begins with the Position and Attitude Data System (PADS), a direct georeferencing platform that fused GPS and inertial sensor measurements via Kalman filtering. Central to that architecture was the Inertial Measurement Unit (IMU), paired with a GPS receiver and an Integration Computer System. This lineage also includes the GPS Inertial Data Simulator (GIDS), a bench-top tool designed to generate real-time GPS and IMU data streams for testing embedded navigation systems without physical movement.

Within this GPS/INS context, the IMU is the component responsible for raw inertial sensing. It typically contains accelerometers and gyroscopes that measure specific force and angular velocity. These measurements are provided at a high data rate, such as 200 Hz, and are integral to a strapdown navigation solution. The IMU does not output position directly; ra

Details

ther, its data is processed by navigation software, often using a Kalman filter, to compute roll, pitch, and heading.

For those asking what is an IMU, it is fundamentally the sensory core of an inertial navigation system. It provides the continuous, high-frequency motion data that, when integrated with GPS observations, yields a complete and accurate attitude and navigation solution.

An inertial measurement unit (IMU) is a self-contained sensor package that measures and reports a vehicle's specific force and angular rate, which are the raw measurements needed to compute acceleration, velocity, and attitude changes without any external references. In essence, an IMU is the sensory core of an inertial navigation system (INS), providing the high-frequency data that a navigation computer integrates to track motion relative to a known starting point. For navigation and avionics engineers, the IMU is not a position sensor itself; it is a dead-reckoning instrument whose outputs must be carefully modeled, filtered, and combined with other data sources to produce a usable navigation solution.

The fundamental principle of an IMU is the measurement of acceleration and rotation in three dimensions. A typical IMU contains three orthogonal accelerometers and three orthogonal gyroscopes (or gyros). The accelerometers measure specific force, which is the vector sum of linear acceleration and the opposite of gravitational acceleration. The gyros measure angular velocity about each axis. By integrating the angular rates, the system can track its orientation; by integrating the specific force measurements after compensating for gravity and rotation, the system can track its velocity and position. This process is purely mechanical and electrical, requiring no external signals, which makes the IMU immune to jamming or signal blockage. However, this self-containment comes at a cost: the integration of small sensor errors leads to unbounded position and attitude drift over time, a fundamental limitation that defines the role of the IMU in a larger navigation architecture.

Because the IMU measures changes rather than absolute states, its errors are the primary driver of navigation performance degradation. These errors are not random noise alone; they include systematic biases, scale factor errors, misalignments, and non-linearities. In high-accuracy systems, these errors are modeled as stochastic processes, often as first-order Gauss-Markov processes, to allow a Kalman filter to estimate and correct them in real time [1]. The state vector for such a filter typically includes the inertial position and velocity, a three-dimensional attitude deviation, and the IMU error states [1]. The error model must capture accelerometer scale factor asymmetry, accelerometer non-linearities, and gyro randomness, as these terms have a significant effect on calibration and alignment performance [3]. The completeness of these models is a major concern; if an error source is not modeled, it cannot be estimated and will corrupt the navigation solution [3].

The physical performance limits of an IMU are defined by its mechanical and optical design. For example, a stable-member IMU platform used in the Space Shuttle was capable of remaining inertial for vehicle rotations of up to 35 degrees per second and angular accelerations of 35 degrees per second squared [2]. This specification indicates the maximum dynamic environment in which the platform could maintain its reference orientation. Beyond these rates, the gimbal or platform mechanics would saturate, and the IMU would lose its alignment. For strapdown systems, which are rigidly mounted to the vehicle, the gyro range must be high enough to measure the full vehicle rotation rate, and the accelerometer range must cover the expected linear accelerations. The choice between a gimbaled (stable-member) and a strapdown design is a fundamental architectural decision that affects size, cost, and the computational burden of the attitude algorithm.

The output of an IMU is not a clean, continuous signal; it is a discrete stream of data that must be synchronized and pre-processed. In practice, the IMU's internal clock is often not synchronized with the host vehicle's bus clock, leading to a time tag error that drifts almost linearly from 0 to 10 milliseconds before resetting [4]. This drift can cause missing or repeated samples, especially when clock noise causes the sampling edges to pass each other [4]. For a system sampling at 100 Hz, a 10-millisecond time tag error represents a full sample period, which is unacceptable for high-bandwidth control loops [4]. Therefore, the navigation computer must implement a time synchronization and interpolation scheme to align the IMU data with other sensor data, such as GPS measurements. The IMU data is typically pre-processed to remove repeated samples and to correct for the time tag drift before it is passed to the navigation filter [4].

The role of the IMU in an integrated navigation system is to provide the high-rate propagation of the state between updates from absolute sensors like GPS. A GPS receiver provides position and velocity at a relatively low rate, often 1 to 10 Hz, and is subject to signal outages and multipath errors. The IMU, by contrast, provides data at a much higher rate, often 100 Hz or more, and is continuous. In a tightly coupled or deeply integrated INS/GPS system, the IMU data is used to propagate the navigation solution forward in time, while the GPS measurements are used to correct the accumulated drift and to estimate the IMU error states. This architecture is known as an integrated navigation system, where the main computer uses the GPS data in the navigation solution together with other sensors and can also aid the GPS receiver with navigation data to improve satellite reacquisition [8]. The IMU error states are estimated during this process, allowing the filter to compensate for biases and scale factor errors in real time [6].

The integration of an IMU into a vehicle's avionics architecture requires careful consideration of data interfaces and fault management. In the Space Shuttle program, the original concept was to make the new strapdown IMU data look like the legacy stable-member IMU data to avoid changes to the flight software, a concept known as "IMU transparency" [5]. However, this approach was abandoned because it made it impossible for Mission Control to identify and take action on suspect gyros and accelerometers [5]. This lesson highlights a critical design principle: the navigation system must have visibility into the individual sensor outputs to perform fault detection and isolation. A redundancy management scheme typically consists of a selection filter and fault detection logic to identify a failed IMU and switch to a healthy one [2]. The raw IMU data, including integrated attitude rates and changes in accumulated sensed velocity, must be available to the flight software for this purpose [5].

The calibration and alignment of an IMU are critical pre-flight procedures that determine the quality of the navigation solution. During a fine-alignment phase, the only measurement used is integrated velocity, and the purpose is to estimate the attitude and the IMU error states [6]. This process is performed while the vehicle is stationary on the pad, and it allows the filter to converge on the initial attitude and to estimate the gyro biases. The performance of this alignment is shown in end-to-end simulations, where the green section indicates coarse align, the orange section indicates fine align, and subsequent colored sections represent ascent, orbital flight, and entry [6]. The quality of the alignment directly affects the subsequent navigation accuracy, as any residual attitude error will cause gravity to be incorrectly compensated, leading to velocity and position errors that grow with time.