⚠️ fleetis a work in progress and is not validated for research, policy or operational use. It is published early so the approach and the comparison againstmalariasimulationcan be looked at and argued with, not so that anyone can rely on its numbers.Concretely, and honestly:
- The API is unstable.
run_simulation_ode()'s signature has already changed once and may change again without deprecation. Known discrepancies against the IBM are open, not resolved. Across the 63-country site-file comparisonfleetruns above the IBM on falciparum clinical and severe incidence, an excess that is not explained. On P. vivax sites it runs further above the IBM, and that is not explained either; and at school age its vivax LM prevalence and clinical incidence run a few per cent high, a limit of the mean field. The current verdict on every claim, with the numbers behind it, is in fleetcheck — which is kept current, unlike any figure quoted here would be.- Severe incidence and anything derived from it (including DALYs) should be treated as indicative only.
- Nothing here has been peer reviewed, and there is no versioned release.
If you need results you can defend today, use malariasimulation.
A fast, deterministic mean-field twin of the malariasimulation individual-based model of Plasmodium falciparum and P. vivax malaria, run on its daily clock: same inputs, same outputs, population-independent.
fleet reproduces the Griffin-style human model, structured by age and biting heterogeneity, coupled to the compartmental mosquito model, and for P. vivax the White-style model with its hypnozoite batches and relapse. Like malariasimulation, it advances one day at a time: every day it applies immunity decay, biting, one competing draw per person between infection and progression, a day of the mosquito model, and deaths and births, in the IBM's order. It is written in odin2 / dust2.
| Human states | S / D / A / U / Tr, plus the post-treatment and chemoprevention prophylaxis chains and the chemoprevention treated phase; for vivax each over 0 to 10 hypnozoite batches, with liver-stage protection after radical cure |
| Immunity | falciparum: four acquired states IB / ICA / ID / IVA and two maternal terms ICM / IVM, algebraic rather than state variables; vivax: IAA / ICA with their within-cell spread and refractory windows, and maternal IAM / ICM |
| Mosquito, per species | E / L / P / Sm / Em / Im, with the IBM's own delay lines for incubation, the EIR and human infectivity |
It exists to give the malariasimulation ecosystem a deterministic, Monte-Carlo-free companion that:
- Takes the same inputs as the individual-based model (IBM).
run_simulation_ode()accepts amalariasimulation::get_parameters()list, with the usualset_*intervention builders layered on, unchanged, for either parasite:get_parameters(parasite = "vivax")runs the vivax model. - Is seeded where the IBM is seeded. Initial conditions come from
malariaEquilibrium, ormalariaEquilibriumVivaxfor vivax, under the same treatment coverage the IBM seeds with. An undisturbed falciparum run's prevalence relaxes off that seed by under half a percent and then holds, while incidence relaxes by a few per cent over the first years, more at high transmission. A vivax run moves further, as the IBM's does: its clinical incidence settles about 20% above the seed and its realised EIR 4 to 17% aboveinit_EIR, taking up to two decades at the lowest EIRs. So burn in before comparing levels, as with the IBM. - Returns malariasimulation's output table. A parameter list gives the columns the IBM would give it, under the same names and with the same meanings, so a post-processing pipeline written for the IBM, postie included, works on a
fleetrun unchanged. - Is fast and population-independent. All compartments are per-capita densities, so a run costs the same whether you model a thousand people or ten million: a 30-year falciparum run takes about 3 s on the default 209-group age grid (0.7 s on 53 groups), a vivax run, with its hypnozoite dimension, about a minute.
Reach for the IBM instead when you need stochastic variation or individual heterogeneity beyond the mean field.
fleet compiles C++ at install time, as do several of its dependencies, so you need a working C++ toolchain first: Rtools on Windows, the Xcode command line tools on macOS, the usual build tools (r-base-dev or equivalent) on Linux.
# install.packages("remotes")
remotes::install_github("pwinskill/fleet")That also installs the GitHub-only hard dependencies (dust2, monty, malariaEquilibrium and malariaEquilibriumVivax), which DESCRIPTION Remotes points at. It installs nothing from Suggests; the examples need two of those:
remotes::install_github(c("mrc-ide/malariasimulation", "mrc-ide/postie"))malariasimulation builds the parameter list and postie post-processes the output.
library(fleet)
# ask for incidence by age as you would of the IBM: fleet renders exactly the
# bands the list asks for (the defaults give 2-10 prevalence and no incidence)
bands <- list(min = c(0, 5, 15) * 365, max = c(5, 15, 100) * 365 - 1)
p <- malariasimulation::get_parameters(list(
clinical_incidence_rendering_min_ages = bands$min,
clinical_incidence_rendering_max_ages = bands$max,
severe_incidence_rendering_min_ages = bands$min,
severe_incidence_rendering_max_ages = bands$max))
# 10-year daily table, in malariasimulation's columns
out <- run_simulation_ode(timesteps = 3650, parameters = malariasimulation::set_equilibrium(p, init_EIR = 20))
# postie-format rates and prevalence, exactly as for an IBM run
postie::get_prevalence(out, diagnostic = "lm")$lm_prevalence_2_10
postie::get_rates(out)[, c("time", "age_lower", "age_upper", "clinical", "severe", "dalys")]Layer interventions with the ordinary malariasimulation builders and re-run; everything is applied automatically from the parameter list, with no extra arguments. A vivax list runs the same way:
p <- malariasimulation::set_drugs(p, list(malariasimulation::AL_params))
p <- malariasimulation::set_clinical_treatment(p, drug = 1, timesteps = 1, coverages = 0.4)
p <- malariasimulation::set_bednets(
p, timesteps = 365, coverages = 0.6, retention = 3 * 365,
dn0 = matrix(0.387, 1, 1), rn = matrix(0.563, 1, 1),
rnm = matrix(0.24, 1, 1), gamman = 2.64 * 365)
out <- run_simulation_ode(timesteps = 3650, parameters = malariasimulation::set_equilibrium(p, init_EIR = 20))
# P. vivax, with radical cure: relapses and hypnozoite carriage come back as columns.
# Radical cure starts on day 1 here, from a seed without it, so these rows are
# still settling; burn in first for a level to compare.
pv <- malariasimulation::get_parameters(parasite = "vivax")
pv <- malariasimulation::set_drugs(pv, list(malariasimulation::CQ_PQ_params_vivax))
pv <- malariasimulation::set_clinical_treatment(pv, drug = 1, timesteps = 1, coverages = 0.4)
out_pv <- run_simulation_ode(timesteps = 3650, parameters = malariasimulation::set_equilibrium(pv, init_EIR = 3))
tail(out_pv[, c("n_relapses", "n_with_hypnozoites", "iaa_mean")])Three exported functions: run_simulation_ode() runs the model, ode_tuning() holds the discretisation settings, and default_age_lower() gives the default graded age grid. Everything else comes off the parameter list. (The ode in two of the names is kept for compatibility: the model is a daily update, not an ODE.)
That question has its own project: fleetcheck, a register of claims about how closely fleet reproduces malariasimulation, each with the criterion that decides it, the value measured against it, and a verdict.
It lives outside this repository on purpose. Evidence kept beside the code it vouches for is evidence the code's author can quietly leave out of date, so this README quotes no comparison figures of its own.
fleetcheck runs its own CI, fails when a verdict and the register disagree, and refuses to tolerate a failing claim that has no written reason.
| Get started | A worked tour: run the model, read the outputs with postie, layer on each intervention, handle seasonality and burn-in. |
| Using fleet well | Where the mean field departs from the IBM and what to do about it: things to do, things to leave alone, results to treat with caution, and what a run costs. |
| fleetcheck | The evidence, as a separate project: a register of claims about agreement with the IBM — core relationships, age structure, demography, interventions, and 63 country site files — each with its criterion, measurement and verdict. |
| Model specification | The formal version: scope, what the state space does and does not carry, and the full daily update. |
| Parameter reference | Every malariasimulation set_*() function argument by argument: what fleet reproduces exactly, what it approximates, and what it rejects. |
Function reference: ?run_simulation_ode, ?ode_tuning, ?default_age_lower. Contributing: CONTRIBUTING.md.
MIT. Copyright (c) 2026 Peter Winskill. Full text: LICENSE.md.