Robot
- 6 state
- 3 control
- 1 variable
- Mayer cost
- free final time
- 3 state box
- 3 control box
- 1 variable box
- 12 boundary
- OptimalControl valid
- OptimalControl, scalarized valid
- JuMP valid
The robot arm motion problem is a benchmark in time-optimal control. It models a robotic arm moving between two points in space, represented as a rigid bar of total length
Mathematical formulation
The problem can be stated as
The horizon is free, as the total transfer time
Qualitative behaviour
The optimal solution exploits the degrees of freedom of the spherical coordinate system to minimise the motion time. Control inputs
Characteristics
Nonlinear, second-order dynamics in spherical coordinates.
Bang–bang control structure in the time-optimal solution.
Serves as a benchmark for direct transcription and NLP solvers in robotic motion planning.
References
Mössner-Beigel, M. (1989). Thesis, Heidelberg University. Introduces the time-optimal robotic arm problem and its theoretical formulation.
Vanderbei, R. J. (2001). Case studies in trajectory optimization: Trains, Planes, and Other Pastimes. Provides numerical strategies for robotic motion planning and discusses control substitution techniques for time-optimal trajectories.
BOCOP examples: Robot arm problem. BOCOP Robot Arm Example Demonstrates the practical implementation using direct transcription and NLP solvers for robotic motion problems.
Parameters
| Parameter | Default value | Unit | Meaning |
|---|---|---|---|
t0 | 0 | – | initial time |
L | 5 | – | length of the arm, and upper bound of ρ |
uρ_l | -1 | – | lower bound of the control uρ |
uρ_u | 1 | – | upper bound of the control uρ |
uθ_l | -1 | – | lower bound of the control uθ |
uθ_u | 1 | – | upper bound of the control uθ |
uϕ_l | -1 | – | lower bound of the control uϕ |
uϕ_u | 1 | – | upper bound of the control uϕ |
ρ_t0 | 4.5 | – | initial length ρ |
θ_t0 | 0 | rad | initial horizontal angle θ |
ϕ_t0 | 0.7853981633974483 | rad | initial vertical angle ϕ |
dρ_t0 | 0 | – | initial derivative of ρ |
dθ_t0 | 0 | rad | initial derivative of θ, per unit of time |
dϕ_t0 | 0 | rad | initial derivative of ϕ, per unit of time |
ρ_tf | 4.5 | – | final length ρ |
θ_tf | 2.0943951023931953 | rad | final horizontal angle θ |
ϕ_tf | 0.7853981633974483 | rad | final vertical angle ϕ |
dρ_tf | 0 | – | final derivative of ρ |
dθ_tf | 0 | rad | final derivative of θ, per unit of time |
dϕ_tf | 0 | rad | final derivative of ϕ, per unit of time |
ρ_l | 0 | – | lower bound of the length ρ |
θ_l | -3.141592653589793 | rad | lower bound of the angle θ |
θ_u | π | rad | upper bound of the angle θ |
ϕ_l | 0 | rad | lower bound of the angle ϕ |
ϕ_u | π | rad | upper bound of the angle ϕ |
tf_l | 0.1 | – | lower bound of the final time |
Every modeling takes other values with the keyword parameters, merged with these: robot(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 | 250 | 9.141126989 | 8 | 1 | yes, within 1.6e-07 |
| OptimalControl | :trapeze | 250 | 9.141126989 | 18 | 1 | yes, within 2.0e-07 |
| OptimalControl, scalarized | :trapeze | 250 | 9.141126989 | 3 | 1 | yes, within 0 |