optimization¶

Optimization of virtual mechanisms: plan the closed loop’s motion and tune the Params.

Modules

`asktell`	Gradient-free tuning of Params by ask and tell: you run the episodes, it picks the next.
`builder`	What a block adds to: variables, constraints and costs, and the system's compiled functions.
`collocation`	Collocation: the whole motion at once, with the closed loop as constraints.
`equilibrium`	Equilibrium: the closed loop at rest, as a static problem.
`horizon`	Moving-horizon estimation: the state that best explains the last readings, as a least squares.
`mpc`	MPC: the plan as a receding-horizon controller of a virtual mechanism's Params.
`nlp`	The assembled nonlinear program, ready for any solver.
`problem`	Problem: a system, the Params to optimize, and the blocks that make the program.
`result`	The outcome of a solve: the planned motion, the values found, and how to apply them.
`search`	Where the references sit, searched on a grid: one reference at a time, the best plan kept.
`shooting`	Shooting: one rollout of the closed loop per interval, joined by continuity constraints.
`solver`	The nonlinear solver: IPOPT and the other presets, and the progress of a solve.
`tank`	TankBudget: the planned changes of the controller's Params stay within a tank's energy.
`terms`	Terms of a problem: the effort, a coordinate to bring to zero, a coordinate to keep in bounds.
`trajectory`	The nodes of a planned motion, as the terms of a problem see them.
`variables`	Decision variables stacked into one vector, each scaled so that the solver sees values near 1.
`virtual`	Virtual states in a plan: a controller's own states, as unknowns advanced as it advances them.