lucosanta/ppaml-vrep-automobile-plugin

V-REP plugin to extract sensor data from a vehicle with Ackermann steering

                      ppaml-vrep-automobile-plugin, v1.0.1

This package provides a data collection plugin for the V-REP robotics simulator
<http://coppeliarobotics.com/>.  The plugin is targeted at data collection
following the model described at
<http://prism2.mem.drexel.edu/~vefa/courses/SLAM/nebotJournal.pdf>, but it may
be useful for other projects as well.

For detailed installation instructions, see the INSTALL file.

For examples of use, see the examples directory.

Once you have installed the library and restarted V-REP, you'll have at your
disposal five new Lua functions:

  - simExtAutomobileInit(string directoryName, number L, number h, number a,
                         number b, number theta0, number max_distance,
                         number max_intensity)
    Initializes the plugin.  Consequently, you must call this at least once
    during your run, and we strongly recommend you do so before collecting any
    data. :)  The parameters are:

      - directoryName: The base directory for the output data.  If the
        directory does not exist, the plugin will create it.  If the directory
        does exist, the plugin will remove anything in it that looks like data
        from a previous run.

      - L, h, a, b, theta0: Model parameters as described in the paper.
        Specifically:

          - L: Distance between front and rear axles.

          - h: Distance between center of rear axle and encoder on left rear
            wheel.  In examples/simple.ttt, this parameter is zero, as no
            encoder is used; instead, we measure position directly at the
            center of the rear axle.

          - a, b: Distance between the rear axle and the lidar sensor,
            expressed as a 2D vector.  The basis vectors point in line with the
            car and forward (for a) and in line with the axle and to the left
            side of the vehicle (for b).

          - theta0: Initial angle of the car, with 0 defined as due east and
            increasing counterclockwise (i.e., in the positive k̂ direction).

      - max_distance, max_intensity: Lidar parameters.  max_distance should be
        set to the distance between the lidar sensors' apertures and their far
        clip planes; in examples/simple.ttt, this is 10.  max_intensity is
        entirely dependent on what code you're using to postprocess the data;
        we've had it set at 32768, but you can pick whatever you want without
        incident.

  - simExtAutomobileRequestNoise(table2 xy, table2 angle, table2 speed,
                                 table2 steeringAngle, table2 intensity,
                                 table2 distance)
    Requests that, in addition to the gathered data, the plugin also generate a
    data set with artificial Gaussian noise added.  Should you wish to use this
    function, we strongly recommend you only call it once per run.  The
    parameters are two-element numeric tables describing the mean and standard
    deviation for the noise applied to each measurement.

  - simExtAutomobileSavePose(number simulationTime, number x, number y,
                             number theta)
    Records a pose (position and angle).  You should call this repeatedly to
    save multiple poses over the course of a simulation.

  - simExtAutomobileSaveControls(number simulationTime, number speed,
                                 number steeringAngle)
    Records measurements taken from control surfaces of the car--speed and
    steering angle of the right front wheel.  (This model has Ackermann
    steering; hence, the two front wheels will generally not be oriented at
    equal angles.)  You should call this repeatedly to save multiple data over
    the course of a simulation.

  - simExtAutomobileSaveLaserPair(number simulationTime, table leftDepthBuffer,
                                  table rightDepthBuffer, table leftImage,
                                  table rightImage)
    Records measurements from the lidar.  You should call this repeatedbly to
    save multiple data over the course of a simulation.  So that you may
    simulate a lidar with a wider angle than V-REP's distance sensor part
    supports, you can pass two inputs, which will be concatenated in the plugin
    before being written to the file.  Your input tables may be of any length,
    but they must be the same length on each call to this function throughout a
    single run--should you fail to maintainf this invariant, the plugin will
    crash with a C++ exception, bringing down V-REP with it.

The output directory tree will look like this:

    output_dir
    ├── ground
    │   ├── slam_control.csv
    │   ├── slam_gps.csv
    │   ├── slam_laser.csv
    │   └── slam_sensor.csv
    ├── noisy
    │   ├── slam_control.csv
    │   ├── slam_gps.csv
    │   ├── slam_laser.csv
    │   └── slam_sensor.csv
    └── properties.csv

The properties.csv file contains run properties written with
simExtAutomobileInit.  The ground subdirectory contains ground truth data; the
noisy subdirectory contains data with additive noise.  In each subdirectory,
you'll find

  - slam_sensor.csv: A "table of contents" file that describes which sensor was
    sampled at what time.

  - slam_gps.csv: All poses saved with simExtAutomobileSavePose.

  - slam_sensor.csv: All data saved with simExtAutomobileSaveControls.

  - slam_laser.csv: All lidar measurements saved with
    simExtAutomobileSaveLaserPair.