This package extends the interfaces defined in `DifferentialDynamicsModels.jl`

to simple car dynamics of the form

The state consists of position and heading in the plane. The control inputs are speed and curvature , or some derivative of each of them (e.g., curvature rate as input ensures trajectories with continuous curvature) with the state and dynamics equation correspondingly augmented. This package exports the type `SimpleCarDynamics{Dv,Dκ} <: DifferentialDynamics`

to represent these dynamics, where `Dv`

and `Dκ`

denote the number of integrators in the speed and curvature control inputs respectively.

In addition to providing these dynamics and a few methods for exact propagation, this package also contains pure Julia implementations of minimum arc length Dubins and Reeds-Shepp steering for a simple car with minimum turning radius `r`

. These implementations aim to be non-allocating and highly performant (e.g., for use in robotic motion planning where computing millions of steering connections in a few seconds may be necessary), and may be accessed through the `SteeringBVP`

interface defined in `DifferentialDynamicsModels.jl`

:

`DubinsSteering(; v=1, r=1)`

returns a`DubinsSteering`

instance which may be called with two`SE2State`

s as input, or any pair of`StaticVector`

s of length 3.`ReedsSheppSteering(; v=1, r=1)`

returns a`ReedsSheppSteering`

instance which may be called as above.

or through specialized functions:

`dubins_length(q0, qf; r=1)`

and`dubins_waypoints(q0, qf, dt_or_N; v=1, r=1)`

give the length and a`Vector`

of states along the optimal Dubins steering curve respectively.`dt_or_N`

may be an`AbstractFloat`

or an`Int`

, corresponding to a desired time spacing`dt`

(with car speed`v`

) or a desired total number of equally spaced waypoints`N`

.`reedsshepp_length(q0, qf; r=1)`

and`reedsshepp_waypoints(q0, qf, dt_or_N; v=1, r=1)`

give the length and a`Vector`

of states along the optimal Reeds-Shepp steering curve respectively.

10/22/2017

almost 2 years ago

9 commits