# google-research/planet

Learning Latent Dynamics for Planning from Pixels

Repository: https://github.com/google-research/planet
Canonical: https://ross.abutalabs.com/products/google-research-planet
Homepage: https://danijar.com/planet
Language: Python
License: Apache-2.0
License Family: permissive
Archived: true
Last push: 2023-03-24T21:57:54+00:00

## Health v2 (maintenance only)
Score: 10/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 2758, "days_push": 1258, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, archived
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1260, forks 218 (observed 2026-08-28T04:04:10.029732+00:00)

## What it is
Open-source implementation of the PlaNet agent, a purely model-based reinforcement learning algorithm that solves control tasks from pixels by planning in a learned latent space. It encodes image histories into compact latent states and predicts future rewards for candidate action sequences to act efficiently with minimal environment interaction.

## Use cases
- train a model-based RL agent on control tasks from images
- reproduce PlaNet results from the ICML 2019 paper
- experiment with latent dynamics models for planning
- benchmark model-based vs model-free RL sample efficiency
- modify latent transition models for RL research
- run RL experiments on DeepMind Control Suite tasks like cheetah_run

## When to choose
- you need a model-based RL baseline that learns from pixels with low sample complexity
- you want to study or extend latent dynamics world models for control
- you are reproducing or building on the PlaNet paper

## When to avoid
- you need a production-ready or actively maintained RL framework
- you want model-free RL algorithms like SAC or PPO
- you need a simple high-level API rather than research code

## Facets
- artifact type: library
- maturity: maintenance
- function: reinforcement-learning, machine-learning, simulation
- domain: reinforcement-learning, machine-learning, artificial-intelligence
- platform: python
- tags: model-based-rl, latent-dynamics, planning, research-code, deep-learning, linux, gpu

## Member repositories
- google-research/planet (main) score 10

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:10.029732+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T05:04:23.689106+00:00, confidence not recorded.
  - readme: https://github.com/google-research/planet (fetched 2026-08-28T04:04:10.029732+00:00, sha ab7da5567b4a)
  - homepage: https://danijar.com/planet (fetched 2026-08-29T12:17:08.000024+00:00, sha 1870e1b69533)
- Data as of 2026-08-30T08:39:29.467469+00:00.