Dead Reckoning Trade Study Reducing Drift with Varying PNT Aiding Sensor Combinations Through Sensor Simulation and Real-World Evaluation
Tracey Young, Alex Stevenson, U.S. Army DEVCOM C5ISR Center; Daniel Jacobs, John Raquet, Jacob Pate, Integrated Solutions for Systems, (IS4S); Jeremy Gray, AFIT ANT Center/HII; Jonathan Accurso, AFIT ANT Center/CAL Analytics
Location: Ballroom D
Date/Time: Tuesday, Jun. 2, 8:55 a.m.
In inertial navigation systems, the pairing of global navigation satellite systems (GNSS) for low-frequency absolute position updates and inertial measurement units (IMUs) for high-frequency dynamics can be a brittle source of navigation in the face of GNSS outages and limited IMU performance. Due to the nature of inertial mechanization, small IMU errors accumulate to result in exponential error growth. In lower size, weight, and power (SWaP) systems, particularly, even brief GNSS outages can be enough to render the mechanized IMU solution useless unless the drift is bounded by other sources.
This presentation compares the results of using various combinations of relatively low-cost inputs to limit the error of the navigation solution in the absence of absolute GNSS positioning. These inputs include differing qualities of various PNT sensors. The measurements from these sensors can constrain and reduce the dead reckoning solution’s error growth from exponential to linear. Additional information can be obtained from these inputs without requiring extra hardware. Information from the existing inputs as well as from known vehicle dynamics can be used to generate estimated measurements, especially during stationary periods.
Results are presented from both simulated and physical sensors. The simulated results were used to assess performance for a variety of system configurations, with a focus on varying cost-performance grades of the PNT sensors, as well as to inform sensor performance requirements prior to the purchase of real sensors. The real-world results were then used to refine the sensor error models used in the filter. The updated models were then fed back into the simulation, validated against real world results, and used to complete assessments of sensor configurations that were unable to be tested with real sensors. Finally, a clear analysis was performed to highlight the tradeoff between cost and performance among the various configurations.