mdp_planning: mdp planning

In cboettig/mdplearning: Bayesian Learning Algorithms for Markov Decision Processes

Description

Simulate MDP under a given policy

Usage

mdp_planning(transition, reward, discount, model_prior = NULL, x0,
  Tmax = 20, observation = NULL, a0 = 1, policy, ...)

Arguments

transition	list of transition matrices, one per model
reward	the utility matrix U(x,a) of being at state x and taking action a
discount	the discount factor (1 is no discounting)
model_prior	the prior belief over models, a numeric of length(transitions). Uniform by default
x0	initial state
Tmax	length of time to simulate
observation	NULL by default, simulate perfect observations
a0	previous action before starting, irrelivant unless actions influence observations and true_observation is not null
policy	a vector of length n_obs, whose i'th entry is the index of the optimal action given the system is in (observed) state i.
...	additional arguments to mdp_compute_policy

Value

a data frame "df" with the state, action and a value at each time step in the simulation

Examples

source(system.file("examples/K_models.R", package="mdplearning"))
transition <- lapply(models, `[[`, "transition")
reward <- models[[1]]$reward

df <- mdp_compute_policy(transition, reward, discount, model_prior = c(0.5, 0.5))
out <- mdp_planning(transition[[1]], reward, discount, x0 = 10,
               Tmax = 20, policy = df$policy)

## Simulate MDP strategy under observation uncertainty
out <- mdp_planning(transition[[1]], reward, discount, x0 = 10,
               Tmax = 20, policy = df$policy, 
               observation = models[[1]]$observation)

cboettig/mdplearning documentation built on May 13, 2019, 2:08 p.m.