Robotic Dinosaur Locomotion

PREHISTORICROBOTICINTELLIGENCE

Open-source platform for building robotic dinosaurs
powered by physics simulation and reinforcement learning

SPECIES: 3
TESTS PASSING: 155
TRAINING STAGES: 3

Core Features

MuJoCo Physics

Accurate dinosaur biomechanics simulation with articulated joints, contact dynamics, and actuator models using MuJoCo.

Reinforcement Learning

PPO and SAC algorithms via Stable-Baselines3 with automated 3-stage curriculum learning: balance, locomotion, behavior.

3 Species

T-Rex (18 actuators), Velociraptor (17 actuators), and Brachiosaurus (22 actuators) — bipedal and quadrupedal gaits.

Open Source

MIT licensed. TOML configs, Gymnasium registration, W&B tracking, Docker support, and Vertex AI cloud training.

See It In Action

Velociraptor learning to balance using PPO reinforcement learning — Velociraptor learning to balance via PPO curriculum training

Training Stage Videos

Velociraptor training progression across each curriculum stage

STAGE 1

Balance

Raptor learning to stand without falling

STAGE 2

Locomotion

Raptor learning to walk and run forward

STAGE 3

Strike

Raptor sprinting and attacking with sickle claws

Project Roadmap

PHASE 0 — v0.2.0

Clean Slate

TOML configs
Gymnasium registration
Developer tooling
Testing improvements

COMPLETE

PHASE 1 — v0.3.0

First Steps

Curriculum manager
W&B tracking
Locomotion metrics
Species training runs

IN PROGRESS

PHASE 2 — v0.4.0

Into the Wild

Domain randomization
Terrain diversity
Turning & steering

PLANNED

PHASE 3 — v0.5.0

Evolution

Custom policy networks
New species
Benchmark suite

PLANNED

PHASE 4 — v0.6.0

The Pack

Multi-agent envs
Cooperative hunting
Predator-prey

PLANNED

PHASES 5-6

Hyperdrive & Sim-to-Real

JAX/MJX backend
Hardware prototype
ROS 2 bridge

PLANNED

Ready to Build Robotic Dinosaurs?

Explore the docs, train your first model, or contribute to the project.

GET STARTED VIEW ON GITHUB