How to Run an Accurate Inertial Navigation Simulation in Less Than 5 Minutes
Just about every gadget that moves, from the Apple iWatch to SpaceX’s Falcon 9 rocket has an IMU ( Inertial Measurement Unit) sensor inside it that helps the device keep track of its location as well as its movements. Yes, the iWatch and Falcon 9 rocket use vastly different accuracy IMUs and process IMU data with vastly different algorithms, but the basic physics and math models underlying each IMU simulation remains the same.
What if there was a convenient on-line way to simulate different IMU related algorithms and quickly understand the impact of key IMU accuracy parameters on performance? Now you can.
ACEINNA has released a full-featured web-based graphical user-interface for its popular open-source GNSS-INS-SIM simulator. Previous posts reviewed how to set-up and use GNSS-INS-SIM from Python directly. Now with the new web-based GUI, the inertial navigation simulator can be used with just a few mouse clicks and in less than 5 minutes of time.
Getting Started
An IMU consists of three axes of acceleration sensing, and three axes of angular rate sensing. An IMU can optionally contain three axes of magnetic sensing, and IMU’s are often coupled with GPS/GNSS receivers for GPS/INS navigation. IMU’s are frequently used for general motion measurement, platform stabilization, attitude measurement, and position tracking during loss of GPS.
To use the new on-line Inertial Navigation simulator, register on ACEINNA’s Developer Website and login. Follow these Step-by-step instructions for a complete tutorial. The remainder of this blog highlights a few of the many features.
Creating an IMU Sensor Model
A sensor model describes both the drift and noise performance of the IMU. These parameters are easily input on-line to create an IMU sensor model. In most cases these parameters are provided by the IMU manufacturer specified as bias stability and random walk on the data-sheet. The rate sensor has an angle random walk, and the accelerometer a velocity random walk. Error models for high-accuracy, medium accuracy, and low-accuracy IMU’s are also provided ready to use. Be sure to give your IMU error model a name and save it before moving to the next step.
Generating a Motion Trajectory
In order to run an algorithm, a motion trajectory is needed. The motion trajectory represents the vehicle or object path versus time, and it is used to generate the reference or “Truth” accelerations and angular rates that are processed. The simulation will generate both a “Truth” result using IMU readings without any errors, as well as an estimated actual result using the sensor model to generate simulated errors. The simplest trajectory of all is static — i.e., no motion which can be used to generate a simulated Allan Variance. More typically a 2-D or 3-D motion trajectory is created, such as the example “drive test” trajectory created with the on-line motion generator and shown below. This trajectory corresponds to the same “left-hand” turn test case discussed in a previous blog, and it is representative of a real-world inertial navigation use case for a self-driving car to navigate an intersection unaided by LIDAR, Cameras, or GPS/GNSS sensors for position determination.
Select an Algorithm and Simulate
With a motion file generated and a sensor model created, you are ready to select an algorithm and run the inertial navigation simulation. Currently the web-based on-line edition of GNSS-INS-SIM works with several pre-defined algorithms including the algorithms which ship on ACEINNA’s line of IMU products. In the future, ACEINNA will provide tools to upload custom algorithms created with ACEINNA’s OpenIMU embedded inertial development tools.
Following the previous blog post, we process the above drive test trajectory with the “Free Integration” algorithm that demonstrates the pure inertial navigation performance of an IMU. The simulation can be run multiple times to generate error statistics. The default is one time.
Results: Allan Variance, Time Series Graphs, Maps, Statistical Analysis and More …
After a short period of time, your simulation results will be ready and you will receive a notification that the simulation has completed. The on-line version of GNSS-INS-SIM provides a number of on-line tools to analyze the simulation data including:
- Statistics Summary of Errors
- Allan Variance Plot (Allan Algorithm)
- Time Series Plots and Comparison/Error Plots
- Map Plots (Position and Reference Position Outputs)
Below are a couple of example plots generated from this simulation run.
Next Steps
Simulation of IMU hardware and Inertial Navigation based algorithms has generally been a tricky and time consuming process. ACEINNA’s Developer Website provides a powerful on-line tool that radically simplifies the process. The simulation tool will be expanded and added to based on user feedback.
In the future, based upon user feedback, the simulation tool will be expanded and further developed. Upcoming posts will detail how to do full GPS/INS simulations and model dynamic GPS outages such as those caused by trees, buildings and other real-world interference sources.