GNSS-INS-SIM

GNSS-INS-SIM is an GNSS/INS simulation project, which generates reference trajectories, IMU sensor output, GPS output, odometer output and magnetometer output. Users choose/set up the sensor model, define the waypoints and provide algorithms, and gnss-ins-sim can generate required data for the algorithms, run the algorithms, plot simulation results, save simulations results, and generate a brief summary.

Contents

• Requirements
• Demos
• Get started
  • Step 1 Define the IMU model
  • Step 2 Create a motion profile
  • Step 3 Create your algorithm
  • Step 4 Run the simulation
  • Step 5 Show results
• Acknowledgement

Requirements

• Numpy ( version>1.10 )
• Matplotlib

Demos

We provide the following demos to show how to use this tool:

Get started

Step 1 Define the IMU model

Step 1.1 Define the IMU error model

IMU error model can be specified in two ways:

Choose a built-in model

There are three built-in IMU models: 'low-accuracy', 'mid-accuracy' and 'high accuracy'.

Manually define the model

imu_err = {
            # gyro bias, deg/hr
            'gyro_b': np.array([0.0, 0.0, 0.0]),
            # gyro angle random walk, deg/rt-hr
            'gyro_arw': np.array([0.25, 0.25, 0.25]),
            # gyro bias instability, deg/hr
            'gyro_b_stability': np.array([3.5, 3.5, 3.5]),
            # gyro bias instability correlation, sec.
            # set this to 'inf' to use a random walk model
            # set this to a positive real number to use a first-order Gauss-Markkov model
            'gyro_b_corr': np.array([100.0, 100.0, 100.0]),
            # accelerometer bias, m/s^2
            'accel_b': np.array([0.0e-3, 0.0e-3, 0.0e-3]),
            # accelerometer velocity random walk, m/s/rt-hr
            'accel_vrw': np.array([0.03119, 0.03009, 0.04779]),
            # accelerometer bias instability, m/s^2
            'accel_b_stability': np.array([4.29e-5, 5.72e-5, 8.02e-5]),
            # accelerometer bias instability correlation, sec. Similar to gyro_b_corr
            'accel_b_corr': np.array([200.0, 200.0, 200.0]),
            # magnetometer noise std, uT
            'mag_std': np.array([0.2, 0.2, 0.2])
          }

Step 1.2 Create an IMU object

imu = imu_model.IMU(accuracy=imu_err, axis=6, gps=False)
imu = imu_model.IMU(accuracy='low-accuracy', axis=9, gps=True)

axis = 6 to generate only gyro and accelerometer data.

axis = 9 to generate magnetometer data besides gyro and accelerometer data.

gps = True to generate GPS data, gps = False not.

Step 2 Create a motion profile

A motion profile specifies the initial states of the vehicle and motion command that drives the vehicle to move, as shown in the following table.

The initial position should be given in the LLA (latitude, longitude and altitude) form. The initial velocity is specified in the vehicle body frame. The initial attitude is represented by Euler angles of ZYX rotation sequence.

Motion commands define how the vehicle moves from its initial state. The simulation will generate true angular velocity, acceleration, magnetic field, position, velocity and attitude according to the commands. Combined with sensor error models, these true values are used to generate gyroscope, accelerometer, magnetometer and GPS output. There is only one motion command in the above table. Indeed, you can add more motion commands to specify the attitude and velocity of the vehicle. You can also define GPS visibility of the vehicle for each command.

Five command types are supported, as listed below.

An example of motion profile

The initial latitude, longitude and altitude of the vehicle are 32deg, 120deg and 0 meter, respectively. The initial velocity of the vehicle is 0. The initial Euler angles are 0deg pitch, 0deg roll and 0deg yaw, which means the vehicle is level and its x axis points to the north.

This command is of type 1. Command type1 directly gives Euler angle change rate and velocity change rate. In this case, they are zeros. That means keep the current state (being static) of the vehicle for 200sec. During this period, GPS is visible.

This command is of type 5. Command type 5 defines attitude change and absolute velocity. In this case, the pitch angle will be increased by 45deg, and the velocity along the x axis of the body frame will be accelerated to 10m/s. This command should be executed within 250sec.

This command is of type 3. Command type 3 defines attitude change and velocity change. In this case, the yaw angle will be increased by 90deg, which is a right turn. The pitch angle is decreased by 45deg. The velocity of the vehicle does not change. This command should be executed within 25sec.

The following figure shows the trajectory generated from the motion commands in the above table. The trajectory sections corresponding to the above three commands are marked by command types 1, 5 and 3.

Step 3 Create your algorithm

algo = allan_analysis.Allan() # an Allan analysis demo algorithm

An algorithm is an object of a Python class. It should at least include the following members:

self.input and self.output

The member variable 'input' tells gnss-ins-sim what data the algorithm need. 'input' is a tuple or list of strings.

The member variable 'output' tells gnss-ins-sim what data the algorithm returns. 'output' is a tuple or list of strings.

Each string in 'input' and 'output' corresponds to a set of data supported by gnss-ins-sim. The following is a list of supported data by gnss-ins-sim.

self.run(self, set_of_input)

This is the main procedure of the algorithm. gnss-ins-sim will call this procedure to run the algorithm. 'set_of_input' is a list of data that is consistent with self.input. For example, if you set self.input = ['fs', 'accel', 'gyro'], you should get the corresponding data this way:

  def run(self, set_of_input):
      # get input
      fs = set_of_input[0]
      accel = set_of_input[1]
      gyro = set_of_input[2]

self.get_results(self)

gnss-ins-sim will call this procedure to get resutls from the algorithm. The return should be consistent with self.output. For example, if you set self.output = ['allan_t', 'allan_std_accel', 'allan_std_gyro'], you should return the results this way:

  def get_results(self):
      self.results = [tau,
                      np.array([avar_ax, avar_ay, avar_az]).T,
                      np.array([avar_wx, avar_wy, avar_wz]).T]
      return self.results

self.reset(self)

gnss-ins-sim will call this procedure after run the algorithm. This is necessary when you want to run the algorithm more than one time and some states of the algorithm should be reinitialized.

Step 4 Run the simulation

step 4.1 Create the simulation object

  sim = ins_sim.Sim(
        # sample rate of imu (gyro and accel), GPS and magnetometer
        [fs, fs_gps, fs_mag],
        # initial conditions and motion definition,
        # see IMU in ins_sim.py for details
        data_path+"//motion_def-90deg_turn.csv",
        # reference frame
        ref_frame=1,
        # the imu object created at step 1
        imu,
        # vehicle maneuver capability
        # [max accel, max angular accel, max angular rate]
        mode=np.array([1.0, 0.5, 2.0]),
        # specifies the vibration model for IMU
        env=None,
        #env=np.genfromtxt(data_path+'//vib_psd.csv', delimiter=',', skip_header=1),
        # the algorithm object created at step 2
        algorithm=algo)

gnss-ins-sim supports running multiple algorithms in one simulation. You can refer to demo_multiple_algorihtms.py for example.

There are three kinds of vibration models: random, sinusoidal and PSD.

For the vibratoin model of the gyro, it is similar as that of the acceleration except that the unit is default to rad/s and 'd' is used to change the unit to be deg/s.

The following example sets random vibration to accel with RMS for x/y/z axis being 1, 2 and 3 m/s^2, respectively, and sets sinusoidal vibration to gyro with frequency being 0.5Hz and amplitude for x/y/z axis being 6, 5 and 4 deg/s, respectively.

    env = {'acc': '[1 2 3]-random',
           'gyro': '[6 5 4]d-0.5Hz-sinusoidal'}

Step 4.2 Run the simulation

sim.run()     # run for 1 time
sim.run(1)    # run for 1 time
sim.run(100)  # run for 100 times

Step 5 Show results

# generate a simulation summary,
# and save the summary and all data in directory './data'.
# You can specify the directory.
sim.results('./data/')

# generate a simulation summary, do not save any file
sim.results()

# plot interested data
sim.plot(['ref_pos', 'gyro'], opt={'ref_pos': '3d'})

Acknowledgement

• Geomagnetic field model https://github.com/cmweiss/geomag/tree/master/geomag
• MRepo http://www.instk.org