Skip to content

TrajectoryOptimization.md

Latest commit

 

History

History
72 lines (59 loc) · 2.67 KB

TrajectoryOptimization.md

File metadata and controls

72 lines (59 loc) · 2.67 KB
Raw

Trajectory optimization

First test case (in Julia)

  1. define inflow conditions
  2. send init request with initial guess
  3. receive trajectory
  4. fly it

Open questions:

Which inflow conditions

My current suggestion would be the struct:

class InflowConditions(TypedDict):
    wind_speed: float                   # in m/s at 6 m height
    wind_direction: float               # in degrees, 0 = North, 90 = East
    profile_law: Literal[0, 1, 2, 3]    # 0=CONST, 1=EXP, 2=LOG, 3=EXPLOG
    # the following fields are optional; the defaults given in the comments are used if omitted
    alpha: NotRequired[float]           # exponent of the wind profile law, default: 0.08163
    z0: NotRequired[float]              # surface roughness [m], default: 0.0002
    turbulence: NotRequired[float]      # in [0, 1], 0 = no turbulence, 1 = full turbulence, default: 0.0

Is that good enough for you, or what shall I add/ change? --> can be implemented, currently only LOG and CONST implemented

I see you defined:

              wind = Dict("model_type" => "tabulated",
                          "heights" => [10.0, 100.0, 300.0],
                          "speeds" => [5.5, 8.0, 9.3]),

I could add that as custom wind profile law. But that would require some discussion first, because it does not really define a profile law. --> Which interpolation would be good? Uwe will investigate.

Which additional parameters are needed for init

I see you passed:

             extras = Dict("sim_parameters" => Dict(
                 "input_depower" => 1.6, "reg_weight" => 1.0,
                 "detect_simple_bounds" => true)))

Shall these parameters be added to InitParams ? We should define for V3 what is zero and what is one for rel_depower. What is the meaning of these parameters? --> Uwe: Check how rel_depower is converted to depower line length in KitePod.jl

Resolved 2026-08-18 (see steering_depower.md): input_depower is l_dp, an absolute power-tape length in metres, and both sides move 5 m of tape per unit of rel_depower — but off different zeros, AWETrim's l_dp = 0.6 + 5*u_p against V3Kite's 0.2 + 5*u_p. The 0.4 m offset is 8 kcu-%. It is not a rescaling that can simply be applied: pinning input_depower at the flown depower leaves the solve infeasible.

I think, all parameters needed for the initialization should be part of the InitParams struct.

Which result are we interested in

  • average force
  • average speed
  • average power
  • RMS path following error in degrees
  • anything else ?

--> Uwe can compare time series data.

Other questions Is an initial guess needed for the trajectory? -> YES Optimizer requires it.