allan_variance

Code for IMU characterization using Allan variance techniques. Plays nicely with ROS bagfiles.

Installing Dependencies

Install pip (if necessary)

sudo easy_install pip

Install allantools (https://github.com/aewallin/allantools)

sudo pip install allantools

Usage

A launch file is provided for using the allan.py script. The script parses a bagfile and performs allan variance on the specified topic and IMU axis.

Parameters

~bagfile_path (string, default: "")

• The global path to the bagile containing static IMU data.

~imu_topic_name (string, default: "/imu")

• The name of the IMU topic within the bagfile.
• Note: topic must be of the sensor_msgs/Imu type.

~axis (int, default: 0)

• Specify which axis/measurement to perform the allan variance. (0 = all six axes, 1 = x-accel, 2 = y-accel, 3 = z-accel, 4 = x-gyro, 5 = y-gyro, 6 = z-gyro).

~sample_rate (int, default: 100)

• The sample rate of the IMU in Hz.

~delta_measurement (bool, default: false)

• 'true' if delta measurements (change in angle/velocity)

• 'false' if rate measurements (rotation rate/acceleration)

~number_of_lags (int, default: 1000)

• The number of lags (tau values) used in calculating the allan variance.

~results_directory_path (string, default: None)

• The path to the directory where the results should be stored.
• If left unspecified, a directory will be created in the workspace (e.g. ~/allan_ws/av_results/allan_accel_x.csv).

To do

• write node that subscribes to IMU topic and calculates Allan variance periodically. Inform user when the AV is within an acceptable amount of error
• use allantools visualization to show a plot of the resulting allan variance and errors
• calculate values for random walk and bias instability based on result
• incorporate other intertial characterization techniques (simulation, Monte Carlo, autocorrelation, etc.)

Maintainer

Dan Pierce ([email protected])