Use solve_ivp() from Scipy with dynamic args

Question

I am trying to use solve_ivp() function to solve for an ODE with state-dependent args. The idea is to update the ODE at each time instants given the current state of the trajectory.

Given the ODE model as the following:

def model(t, y, arg1, arg2):
    '''
    Some Dynamics model that includes arg1 and arg 2
    '''
    return dy_dt

The arg1 and arg2 are calculated using some other function:

def gen_args(y):
    '''
    Calculate arg1 and arg2 based on the state y
    '''
    return arg1, arg2

Then I want to integrate over the ODE model:

scipy.integrate.solve_ivp(model, t_span, y0, method='RK45', t_eval=None, dense_output=False, events=None, vectorized=False)

What would be a nice way to implement this?

I have checked similar questions like here and here, But all the args are external constants.

score 3 · Accepted Answer · answered Apr 26 '19 at 14:06

3

A simple solution is to call gen_args in model:

def model(t, y):
    arg1, arg2 = gen_args(y)
    dy_dt = ...
    return dy_dt

answered Apr 26 '19 at 14:06

Warren Weckesser

110,654
19
194
214

score 0 · Answer 2 · answered Feb 17 '20 at 18:08

Adding on to @Warren Weckesser answer

It makes very less sense to include any state and (or) time-dependent arguments in the args parameter of 'solve_ivp'. They should be included in the 'fun' parameter of 'solve_ivp' itself. Including any state and (or) time-dependent arguments analogous to this answer inspired by @Warren Weckesser is not possible. The args parameter in

solve_ivp( fun, t_span, y0, method='RK45', t_eval=None, dense_output=False, events=None, vectorized=False, args=None, **options, )

can only take in a tuple of constant arguments (for example floats, integers, etc)

Use solve_ivp() from Scipy with dynamic args

2 Answers2