Skip to contents

Simulate Epidemic Dynamics with Model Predictive Control

Epi_MPC_run_wd.Rd

This function simulates an epidemic under model predictive control (MPC) using pre-defined parameters. It dynamically evaluates actions to maximize expected rewards and updates epidemic states accordingly.

Usage

Epi_MPC_run_wd(
  episimdata,
  episettings,
  epi_par,
  noise_par,
  actions,
  pred_days,
  n_ens = 100,
  start_day = 1,
  ndays = nrow(episimdata),
  R_est_wind = 5,
  pathogen = 1,
  susceptibles = 1,
  delay = 0,
  ur = 0,
  r_dir = 1,
  N = 1e+06
)

Arguments

episimdata

A data frame containing simulation data. It should include columns such as "C" (cases), "I" (infected individuals), "Re" (effective reproduction number), "S" (susceptible individuals), "Deaths", and "Lambda" total infectiousness.

epi_par

A data frame containing epidemiological parameters for various pathogens. It should have the following columns: "R0" (basic reproduction number), "gen_time" (generation time), "gen_time_var" (variance of generation time), "CFR" (case fatality rate), "mortality_mean" (mean mortality delay), and "mortality_var" (mortality delay variance).

noise_par

A data frame containing noise parameters for under-reporting and reporting delays:

  • "repd_mean": Mean of the reporting delay.

  • "del_disp": Dispersion parameter for the reporting delay distribution.

  • "ur_mean": Mean for the under-reporting rate.

  • "ur_beta_a": Alpha parameter of the Beta distribution for under-reporting.

actions

A data frame containing control actions. It should include columns like "R_coeff".

pred_days

An integer specifying the number of days to predict ahead in reward calculation.

n_ens

An integer specifying the number of ensemble runs for expected reward calculation. Defaults to 100.

start_day

An integer specifying the start day of the simulation. Defaults to 1.

ndays

An integer specifying the total number of simulation days. Defaults to the number of rows in episimdata.

R_est_wind

An integer specifying the window size for reproduction number estimation. Defaults to 5.

pathogen

An integer or string identifying the pathogen to extract corresponding parameters. Defaults to 1.

susceptibles

A binary value (0 or 1) indicating whether to update the number of susceptibles. Defaults to 1.

delay

A binary value (0 or 1) indicating whether to simulate reporting delays. Defaults to 0.

ur

A binary value (0 or 1) indicating whether to simulate under-reporting. Defaults to 0.

r_dir

An integer specifying the reproduction number adjustments:

  • 1 for direct Re.

  • 2 for logistic adjustments.

  • 0 for using the generation time distribution.

N

A numeric value representing the total population size. Defaults to 1e6.

Value

A data frame containing updated simulation data with computed reproduction numbers, estimated policies, daily infection incidents, cases, deaths, and other epidemic metrics.

Details

The function employs a model predictive control strategy where actions are evaluated periodically (based on a refresh interval, rf). Expected rewards are computed using the Epi_pred_wd function over a specified prediction horizon, and the optimal action is selected. The simulation incorporates effects of under-reporting, reporting delays, and susceptible depletion based on the provided parameters.

Examples

# Example epidemiological data
episimdata <- data.frame(C = c(0, 10), I = c(5, 8), Re = c(1.5, NA), S = c(1000, 990), Deaths = c(0, 1))
epi_par <- data.frame(
  R0 = 2.5, gen_time = 5, gen_time_var = 1,
  CFR = 0.02, mortality_mean = 14, mortality_var = 2
)
noise_par <- data.frame(
  repd_mean = 2, del_disp = 1.5, ur_mean = 0.8, ur_beta_a = 2
)
actions <- data.frame(R_coeff = c(0.9, 1.1))
updated_data <- Epi_MPC_run_wd(
  episimdata = episimdata, epi_par = epi_par, noise_par = noise_par,
  actions = actions, pred_days = 10, n_ens = 50, start_day = 1,
  ndays = 20, R_est_wind = 5, pathogen = 1, susceptibles = 1,
  delay = 0, ur = 0, r_dir = 1, N = 1e6
)
#> Error in Epi_MPC_run_wd(episimdata = episimdata, epi_par = epi_par, noise_par = noise_par,     actions = actions, pred_days = 10, n_ens = 50, start_day = 1,     ndays = 20, R_est_wind = 5, pathogen = 1, susceptibles = 1,     delay = 0, ur = 0, r_dir = 1, N = 1e+06): argument "episettings" is missing, with no default