Skip to content

Simulation modes

There are two primary modes of operation: batch and real-time.

Batch mode

In batch mode, which is the default, all inputs are supplied to the simulate() method of a given sensor and all outputs are returned. This is suitable for open loop simulations, post-processing recorded data, Monte Carlo analysis and similar applications. It is the most efficient simulation method since it allows for optimization of the underlying math operations. See the Aircraft simulation and Driving logs examples.

Real-time mode

In real-time mode, inputs are supplied incrementally and outputs are returned incrementally. The sensor must be created with the inputs mode="real-time" and max_duration parameters set. The max_duration parameter should be an upper bound on the duration of the simulation (in seconds). It is used to precompute time correlated noise terms. This mode is suitable for closed-loop simulations where the sensor outputs form the input to other processes, e.g. motion control, sensor fusion, etc.

Simulation duration

If the simulation duration is not well known ahead of time it is safe and efficient to set max_duration to a large value (up to several hours). If the value is exceed an IndexError will be raised.

gyro = Gyro(model, specification, mode="real-time", max_duration=300.0)

accelerometer = Accelerometer(model, specification, mode="real-time", max_duration=300.0)

imu = IMU(model, specification, mode="real-time", max_duration=300.0)

In real-time mode, each call to the simulate() method must have sufficient inputs to create an output. For example, simulating gyro data from attitude (orientation) inputs requires differencing the attitude input to calculate angular rate. In an incremental simulation pair-wise attitude inputs must be supplied.

An efficient strategy for accumulating all outputs of a simulation are to store the results in a list (or other iterable container) and combine them at the end. See the Robot arm example.

Sample rate simulation

In real-time mode the simulate_sample_rate option of the model is ignored and users should ensure that the incremental inputs are supplied at the appropriate sample rates. No interpolation or resampling is applied.