stack-of-tasks / ddp-actuator-solver

Differential Dynamic Programming Solver

DDP actuator solver

iLQRSolver.py is a library used to compute a sequence of control to reach a specified point with a specified system and cost function.

In order to use the solver, you have to implement a class which describe the dynamic model of the system and another which describe the cost function you will use to solve your problem. The attributes and methods need to have the same names and return the same things.

dynamic model

This class is the one which describe the system you want to command. It needs to have particular attributes and methods.

attributes

Two attributes are mandatory for the solver to work.

self.stateNumber
self.commandNumber

stateNumber attribute allow the solver to know how many variable your state contains. commandNumber specified the number of commands your system have. These 2 attributes are supposed to be unsigned integer. Both these attributes make the Solver as generic as possible.

methods

Two methods are mandatory for the solver to work.

def computeNextState(self,dt,X,U):
    ''' code here '''
    return nextX

def def computeAllModelDeriv(self,dt,X,U):
    ''' code here '''
    return lx,lxx,lu,luu,lux,lxu

$$\begin{aligned} lx &=& \partial \end{aligned}$$

The solver class need to be instantiated with the dynamic model of the system and the cost function you want to apply.

solver = ILQRSolver(model, costFunction)