Glider
- 4 state
- 1 control
- 1 variable
- Mayer cost
- free final time
- 2 state box
- 1 control box
- 1 variable box
- 7 boundary
- OptimalControl valid
- OptimalControl, scalarized valid
- JuMP valid
The hang glider problem is a classical benchmark in optimal control. It consists of steering a hang glider from an initial horizontal position and altitude to a target altitude while maximising the horizontal distance travelled. The glider dynamics incorporate lift, drag, gravity, and the effect of a thermal updraft. The control variable is the lift coefficient
Mathematical formulation
with
and
where
Qualitative behaviour
The optimal trajectory exploits the thermal updraft to maximise horizontal distance.
The lift coefficient
balances altitude loss and horizontal progression. The dynamics are nonlinear due to the coupling of lift, drag, and relative velocity in the thermal.
Horizontal velocity is maintained positive, and the trajectory respects the control and state constraints.
Characteristics
Nonlinear dynamics.
Widely used as a benchmark for trajectory optimisation and direct transcription methods in nonlinear optimal control.
References
Dolan, E. D., & More, J. J. (2004). Benchmarking Optimization Software with COPS 3.0. Argonne National Laboratory. PDF Includes the hang glider problem in the COPS 3.0 benchmark collection with problem formulation and solver comparisons.
PSOPT Example: Hang Glider Problem. psopt.net/list-of-examples Demonstrates a practical implementation of the hang glider optimal control problem in PSOPT.
PSOPT GitHub Repository: Hang Glider Example. github.com/PSOPT/psopt/blob/master/examples/glider/glider.cxx Source code example for testing direct transcription and NLP solvers with the hang glider problem.
Parameters
| Parameter | Default value | Unit | Meaning |
|---|---|---|---|
t0 | 0 | s | initial time |
x_t0 | 0 | m | initial horizontal position |
y_t0 | 1000 | m | initial altitude |
y_tf | 900 | m | final altitude |
vx_t0 | 13.23 | m/s | initial horizontal velocity |
vx_tf | 13.23 | m/s | final horizontal velocity |
vy_t0 | -1.288 | m/s | initial vertical velocity |
vy_tf | -1.288 | m/s | final vertical velocity |
u_c | 2.5 | m/s | maximum velocity of the updraft of the thermal |
r_t0 | 100 | m | radius of the thermal |
m | 100 | kg | mass of the glider |
g | 9.81 | m/s² | gravitational acceleration |
c0 | 0.034 | – | zero-lift drag coefficient |
c1 | 0.069662 | – | induced drag coefficient, D ∝ c0 + c1 cL² |
S | 14 | m² | wing area |
ρ | 1.13 | kg/m³ | air density |
cL_min | 0 | – | lower bound of the lift coefficient cL |
cL_max | 1.4 | – | upper bound of the lift coefficient cL |
tf_l | 0 | s | lower bound of the final time |
x_l | 0 | m | lower bound of the horizontal position |
vx_l | 0 | m/s | lower bound of the horizontal velocity |
Every modeling takes other values with the keyword parameters, merged with these: glider(backend; parameters=(t0=…,)).
Reference objective
The best objective among the solves that converged, in each modeling, on the default grid of the modeling: the modelings are independent, each has its own reference. See how the references are computed.
| Modeling | Scheme | Grid | Reference objective | Solves | Distinct solutions | Configurations agree |
|---|---|---|---|---|---|---|
| JuMP | :trapeze | 500 | -1247.97843 | 8 | 1 | yes, within 1.0e-10 |
| OptimalControl | :trapeze | 500 | -1247.97843 | 18 | 1 | yes, within 1.0e-10 |
| OptimalControl, scalarized | :trapeze | 500 | -1247.97843 | 3 | 1 | yes, within 4.6e-14 |