Inertial Navigation and Dead Reckoning

Accelerometer Gyroscope Principles

The Mechanics of Linear Acceleration Detection

At the absolute core of autonomous flight in non-permissive or signal-denied environments is the ability to measure physical forces acting on the aircraft without relying on any external transmissions. This internal sensory feedback is handled by the Inertial Measurement Unit, which serves as the physical sensory foundation for the flight control system. The primary component of this unit is the accelerometer triad, consisting of three independent sensors mounted perpendicular to one another, aligned precisely with the longitudinal, lateral, and vertical axes of the aircraft's body frame.

To understand how these sensors operate, we must look at the underlying physics of Micro-Electro-Mechanical Systems, commonly known as MEMS. On a microscopic level, a MEMS accelerometer consists of a tiny silicon proof mass suspended by delicate, flexible silicon springs. When the aircraft experiences a change in linear velocity along any of its three axes, the physical inertia of this proof mass causes it to deflect from its nominal, resting position. On either side of the moving mass are microscopic, interdigitated silicon fingers that act as electrical capacitors. As the proof mass moves, the physical distance between these capacitive fingers shifts. This shifting distance alters the overall electrical capacitance of the circuit in direct proportion to the physical acceleration experienced by the aircraft, translating mechanical movement into a precise digital voltage stream.

However, a fundamental principle of inertial navigation is that accelerometers do not measure true kinematic acceleration relative to the Earth. Instead, they measure specific force. Specific force is defined as the non-gravitational acceleration experienced by the sensor relative to a state of free-fall.

When an aircraft is resting completely motionless on a runway, it is accelerating at zero meters per second squared relative to the Earth's surface. Yet, the internal vertical accelerometer will register an upward specific force of exactly one vertical gravity. This occurs because the physical structure of the runway is actively pushing upward against the aircraft's landing gear to counteract the pull of gravity.

To calculate the true physical movement of the aircraft through space, the autopilot's flight computer must use its real-time orientation estimates to project where gravity is pointing relative to the aircraft's current attitude. It must then computationally subtract that constant downward gravitational force from the raw specific force measurements to isolate the true, kinematic acceleration of the vehicle.

The Mechanics of Angular Velocity Measurement

To perform this vital gravity subtraction, the flight computer must track the aircraft's exact orientation in space at every microsecond. This task is handled by the second primary component of the Inertial Measurement Unit: the gyroscope triad. Modern tactical-grade autopilots utilize solid-state MEMS gyroscopes, which completely replace the heavy, delicate spinning wheels of historical aviation with microscopic silicon elements.

Rather than relying on angular momentum, MEMS gyroscopes utilize a physical phenomenon known as the Coriolis effect. Within the sensor housing, microscopic silicon structures are kept in a state of continuous, high-frequency physical vibration. When the aircraft begins to rotate, whether pitching, rolling, or yawing, the Coriolis effect exerts a lateral, perpendicular force on these vibrating structures.

This lateral force causes a microscopic physical displacement of the vibrating mass, which is detected via capacitive finger arrays. The sensor converts this physical displacement into an electrical signal that is directly proportional to the aircraft's angular velocity, typically expressed in degrees per second.

By placing three of these sensors in an orthogonal arrangement, the autopilot gathers a continuous, high-rate stream of data tracking the exact rate of rotation around the aircraft's longitudinal, lateral, and vertical axes.

Compass Alignment and Magnetometer Integration

While accelerometers and gyroscopes form the core of inertial tracking, they are assisted by a third sensor type: the magnetometer. Magnetometers are designed to measure the strength and direction of the Earth’s natural magnetic field along the three physical axes of the aircraft.

Although magnetometers are not inertial sensors, they serve as a critical compass reference. Gyroscopes are excellent at tracking rapid, high-frequency rotations, but they cannot determine absolute direction on their own. The magnetometer provides a stable, low-frequency link to Magnetic North. This absolute direction acts as a vital reference point that prevents the gyroscopic tracking loops from losing their alignment around the vertical yaw axis over time.

Dead Reckoning Navigation

Core Concepts of Kinematic State Propagation

When an autopilot loses its connection to global positioning satellites, it can no longer determine its position by listening to external coordinates. Instead, the flight control system must rely on dead reckoning. Dead reckoning is the process of estimating the aircraft's current location by continuously projecting forward from a last known starting point using estimated speed, time, and heading vectors.

To execute this process, the flight computer maintains a high-frequency computational loop that integrates the motion data gathered by the IMU. The autopilot takes the high-rate angular velocity readings from the gyroscopes and continuously integrates them over time to maintain an accurate orientation map of the aircraft relative to the Earth.

Simultaneously, the computer takes the raw acceleration measurements from the body of the aircraft and rotates those vectors into a stable local geographic coordinate system, such as the North-East-Down (NED) reference frame. Once rotated, the local gravity value is subtracted, isolating the true physical acceleration of the aircraft.

By integrating this kinematic acceleration over elapsed time, the computer derives the vehicle's current velocity. By integrating that velocity a second time, the autopilot calculates the updated three-dimensional coordinates of the UAS.

Navigating Across Aerospace Coordinate Frames

To perform these kinematic projections, the autopilot must manage transformations across multiple reference frames:

• The Body Frame: This coordinate system is physically locked to the aircraft's airframe, with its origin centered at the vehicle's center of gravity. Its forward axis points straight through the nose, its lateral axis points out along the right wing, and its vertical axis points downward through the belly.
• The Local Level Frame (NED): This is a localized geographical frame that moves along with the aircraft but remains aligned with the Earth's surface. Its axes point directly toward geodetic North, East, and downward along the local vertical.
• The Earth-Centered Earth-Fixed Frame: This frame is fixed to the center of the Earth and rotates along with the planet, making it the ideal system for mapping global coordinates like latitude and longitude.

Attitude Representation and the Elimination of Gimbal Lock

To rotate forces from the aircraft's twisting body frame into the stable local NED frame, the flight control system must maintain a continuous representation of orientation. Historically, aviation systems described orientation using Euler Angles (roll, pitch, and yaw). However, Euler angles possess a fatal geometric flaw known as gimbal lock.

Gimbal lock occurs during extreme vertical maneuvers, such as a steep nose-down dive or a vertical climb, where two of the three physical rotation axes align in the same spatial plane. This alignment collapses a mathematical degree of freedom, causing the coordinate transformation software to lock up and freeze the attitude tracking system.

To prevent this catastrophic failure mode in safety-critical autopilots, aerospace engineers utilize Quaternions. A quaternion is a four-parameter mathematical representation of orientation that models any attitude change as a single, clean rotation around a calculated spatial axis.

Because quaternions do not rely on consecutive rotational steps, they are completely free of geometric singularities and mathematical dead-ends. This ensures that no matter how violently an aircraft stalls, spins, or recovers, the attitude tracking loops remain perfectly stable and uninterrupted.

Sensor Drift Accumulation

The Mathematical Reality of Integration Cascades

While an Inertial Navigation System is highly responsive and completely immune to external jamming, it has a fundamental vulnerability: exponential drift. Because dead reckoning relies on integrating physical measurements over time, any tiny imperfection, noise, or error in the raw sensor readings is multiplied and accumulates rapidly in the position estimate.

This rapid drift is caused by the mathematical nature of the integration cascade. If a sensor contains a small, constant offset error, known as a bias, this error behaves predictably but destructively over time.

When a constant accelerometer bias is integrated once to compute velocity, the velocity error grows linearly with time. When integrated a second time to calculate position, this linear velocity error transforms into a position error that expands quadratically (proportional to the square of time).

The effect of a gyroscope bias is even more destructive. A constant error in a gyroscope causes the calculated orientation to tilt away from reality, growing linearly with time. Because of this tilt, the flight computer miscalculates the direction of gravity, projecting a false acceleration component into the horizontal tracking channels.

When this false acceleration is integrated twice, the resulting position estimation error expands cubically (proportional to the cube of time). Within a few minutes of standalone operation, a high-quality MEMS IMU operating without external corrections can drift by several kilometers, leaving the aircraft completely lost.

Thermal Sensitivity and Mechanical Noise in MEMS Substrates

Beyond simple constant biases, solid-state MEMS sensors are highly sensitive to environmental changes, particularly temperature fluctuations. As the flight computer runs and warms up during a mission, the changing internal temperature causes physical thermal expansion and structural stress within the microscopic silicon elements of the IMU.

This thermal stress shifts the sensor's calibration curves, causing the accelerometer and gyroscope biases to drift unpredictably in real time. IMUs undergo extensive factory calibration across a broad range of operating temperatures. This process establishes a precise thermal model that enables compensate for predictable bias shifts as environmental conditions change.

Kalman Filter Optimization

The Statistical Engine of Modern Avionics

To prevent the rapid collapse of the dead reckoning loop, autonomous flight control systems utilize a real-time statistical supervisor known as an Extended Kalman Filter (EKF). The EKF is a high-performance software engine designed to estimate the true state of the aircraft by balancing the strengths and weaknesses of different sensors.

The EKF operates by tracking both the estimated state vector of the aircraft (including position, velocity, and orientation) and the hidden error characteristics of the physical sensors (such as accelerometer and gyroscope biases). The filter runs a continuous, two-step cyclical process: Prediction and Correction.

The Prediction Step: Blind High-Rate Propagation

The Prediction Phase runs at the high sampling frequency of the IMU (typically hundreds of times per second). The EKF uses the raw acceleration and rotation data to blindly project the aircraft's state forward in time.

Because the filter is aware of the sensor noise and thermal drift inherent to the hardware, its internal estimate of uncertainty, represented by a mathematical covariance matrix, grows larger with every microsecond that passes without an external reference update.

The Correction Step: Bounding Drift via Aiding Networks

The Correction Phase is activated whenever an independent aiding sensor provides a fresh measurement. This aiding sensor could be an optical camera tracking ground features, a downward-pointing altimeter, or a visual-inertial odometry engine.

The EKF compares what the alternative sensor actually measures with what the inertial engine predicted it should see. The difference between these two values is known as the innovation residual.

The filter analyzes this residual, calculates its own confidence levels, and derives a statistical weighting factor called the Kalman Gain. The EKF then uses this Kalman Gain to execute two critical corrections:

• It corrects the accumulated drift in the aircraft's position, velocity, and orientation states, instantly shrinking its uncertainty covariance bounds back to a tight, high-confidence level.
• It updates its internal estimation of the IMU’s hidden sensor biases.

By continuously estimating and subtracting these physical sensor biases in real time, the EKF prevents the SINS loop from drifting quadratically or cubically during subsequent prediction phases. This continuous feedback loop ensures that the core dead reckoning engine remains stable, accurate, and reliable, even during long-term GNSS blackouts.