ScenarioRunner

Overview¶

The ScenarioRunner class is a utility designed for testing IMU preintegration and factors. It takes a ground truth trajectory defined by a Scenario object, along with IMU parameters (PreintegrationParams or PreintegrationCombinedParams) and a specified IMU bias, and simulates the measurements an IMU would produce while following that trajectory.

Key capabilities include:

• Calculating the ideal specific force and angular velocity at any time t based on the scenario’s motion.
• Generating noisy IMU measurements by adding the specified bias and sampling from the noise models defined in the parameters.
• Integrating these simulated measurements over a time interval T to produce a PreintegratedImuMeasurements or PreintegratedCombinedMeasurements object.
• Predicting the final state based on an initial state and a preintegrated measurement object.
• Estimating the covariance of the preintegrated measurements or the prediction error via Monte Carlo simulation (useful for verifying the analytical covariance propagation).

Measurement Simulation¶

• actualAngularVelocity(t): Returns the true angular velocity $\omega_b(t)$ from the Scenario.
• actualSpecificForce(t): Calculates the true specific force (what an ideal accelerometer measures) in the body frame. This is the body-frame acceleration $a_b(t)$ minus the body-frame representation of gravity $g_b(t)$:
  $$f_b(t) = a_b(t) - R_{bn}(t) g_n = R_{bn}(t) (a_n(t) - g_n)$$

  (1)
  where $g_n$ is the gravity vector defined in PreintegrationParams.
• measuredAngularVelocity(t): Adds bias and sampled noise to actualAngularVelocity(t).
  $$\omega_{measured} = \omega_b(t) + b_g + \eta_{gd}$$

  (2)
  where $b_g$ is the gyro bias and $\eta_{gd}$ is sampled discrete gyro noise.
• measuredSpecificForce(t): Adds bias and sampled noise to actualSpecificForce(t).
  $$a_{measured} = f_b(t) + b_a + \eta_{ad}$$

  (3)
  where $b_a$ is the accelerometer bias and $\eta_{ad}$ is sampled discrete accelerometer noise.

Key Functionality / API¶

• Constructor: ScenarioRunner(scenario, params, imuSampleTime=0.01, bias=ConstantBias()): Takes the scenario, IMU parameters, sample rate, and the true bias to apply to measurements.
• Measurement Methods: actual...(), measured...() as described above.
• integrate(T, estimatedBias, corrupted=True): Simulates measurements at imuSampleTime intervals for a duration T, optionally adding noise (corrupted=True), and integrates them into a PIM object using estimatedBias as the biasHat for the PIM.
• predict(pim, estimatedBias): Uses the provided PIM (which contains $\Delta R, \Delta p, \Delta v$) and an estimatedBias to predict the final NavState starting from the scenario’s initial state at $t=0$.
• estimateCovariance(T, N, estimatedBias): Performs Monte Carlo simulation: runs integrate N times with different noise samples, calculates the predicted state for each, and computes the sample covariance of the 9D tangent space difference between the Monte Carlo predictions and the mean prediction. Used to verify pim.preintMeasCov().
• estimateNoiseCovariance(N): Samples noise N times and computes the sample covariance, to verify the noise samplers themselves.

There is also a CombinedScenarioRunner inheriting from ScenarioRunner that works with PreintegrationCombinedParams and PreintegratedCombinedMeasurements.

Usage Example¶

Using the AcceleratingScenario from the Scenario example to generate measurements and verify prediction.

Source¶

• ScenarioRunner.h
• ScenarioRunner.cpp