Reference Trajectory Factory

This page documents the reference trajectory factory, which acts as an interface between the motion planner and the controller. More precisely, it extracts the required snippet for control from the reference trajectory (provided by the motion planner). For most controller, this snippet is a single state or output at a given point in time; for model predictive control (MPC) the duration of this snippet corresponds to the prediction horizon.

Constructor

ref_traj = ReferenceTrajectoryFactory(
        delta_t_controller,
        sidt_factory,
        mpc_horizon,
        t_look_ahead
)

with the parameters

Parameter	Variable type and default value	Description
`delta_t_controller`	`float`	Sampling time of the controller.
`sidt_factory`	`StateInputDisturbanceTrajectoryFactoryInterface`	Required for sampling the reference trajectory in between samples of the motion planner. The `sidt_factory` is required for converting states/control inputs to the correct dataclass, see the vehicle model documentation.
`mpc_horizon`	`int = 0`	(Discrete) Prediction horizon of an MPC. (Only required for MPC).
`t_look_ahead`	`float=0.0`	Look-ahead time.

Setting the Reference Trajectory

The state and input reference trajectory state_ref and input_ref, respectively, starting at time t_0 can be set or updated as follows:

ref_traj.set_reference_trajectory(
        state_ref, 
        input_ref, 
        t_0
)

Note that only one factory instance is required to handle both the state and input reference trajectories.

Querying the Reference Trajectory

To extract the required snippet from the reference trajectory, only the current (continuous) time t_now \(\geq\) t_0 has to be provided.

    x_ref, u_ref = ref_traj.get_reference_trajectory_at_time(
        t_now
    )

Since the prediction horizon of the MPC or potential look-ahead times are provided upon initialization, the correct duration and initial time of the snippet are determined internally.

For delta_t_controller = 0.1 and mpc_horizon = 20, x_ref and u_ref are the snippet of state_ref and input_ref from t_now \(\leq t \leq\) t_now \(+ 2\text{s}\) (sampled every delta_t_controller seconds).
For a PID controller with look-ahead t_look_ahead \(=0.5\text{s}\), x_ref and u_ref are the point of state_ref and input_ref at \(t=\)t_now\(+ 0.5\text{s}\).