sim_plus: sim_plus

In boettiger-lab/pomdpplus: Planning and Learning in Uncertain Systems

Description

sim_plus

Usage

sim_plus(models, discount, model_prior = NULL, state_prior = NULL, x0,
  a0 = 1, Tmax, true_model, alphas = NULL, model_names = NULL, ...)

Arguments

models	a list of lists, each of which gives the transition matrix, observation matrix and reward matrix for the model considered
discount	discount rate
model_prior	Prior belief assigned to each model. uniform by default.
state_prior	Prior belief that system is in state x_i and model m_i
x0	starting state of the system
a0	initial action (used to make the first observation, only relevant if observation depends on action)
Tmax	number of time steps to simulate
true_model	a list of the transition matrix, observation matrix, and reward matrix used to simulate draws.
alphas	the alpha vectors for each model, as provided from sarsop_plus, which will otherwise be run each time if not provided.
model_names	vector of identifying names for each model. If none are provided, model posterior columns will be named V1, V2, etc.
...	additional options to sarsop::sarsop, if alphas are not provided

Value

a list with elements (1) df a data.frame with the time, state, observation, action, and value at each timestep, (2) model_posterior, a data.frame of Tmax rows by n_models columns giving the evolution of the belief over models, and (3) state_posterior, a data.frame of Tmax rows by n_states columns giving the evolution of the belief over states

Examples

## Not run: 
source(system.file("examples/K_models.R", package="pomdpplus"))
alphas <- sarsop_plus(models, discount, precision = .1)
out <- sim_plus(models = models, discount = discount,
                x0 = 6, a0 = 1, Tmax = 10,
                true_model = models[[2]], alphas = alphas)

## End(Not run)

boettiger-lab/pomdpplus documentation built on May 24, 2019, 3:05 a.m.