Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/asimovinc/asimov-v0/llms.txt

Use this file to discover all available pages before exploring further.

​
Introduction

Asimov v0 includes a comprehensive physics simulation model built for MuJoCo, a high-performance physics engine widely used in robotics research and reinforcement learning. The simulation provides a digital twin of the physical robot, enabling algorithm development, controller testing, and gait optimization before deployment to hardware.
The simulation model is designed to closely match the physical robot’s dynamics, including accurate inertial properties, joint limits, and collision geometry.

​
What’s Included

The Asimov simulation model provides:

Complete Kinematic Chain

12 DOF bipedal leg system with pelvis, hips, knees, ankles, and articulated toes

Accurate Physics

Realistic inertial properties, joint damping, and friction parameters from CAD models

Collision Detection

Optimized capsule-based collision geometry for efficient contact simulation

Sensor Suite

IMU (gyroscope, accelerometer, orientation) and contact sensors

​
Key Features

  • Dual Mesh System: Separate visual and collision meshes for rendering performance
  • Articulated Toe Joints: Spring-loaded toe mechanisms with 4.5 Nm/rad stiffness
  • Contact Exclusions: Self-collision prevention between adjacent body segments
  • Free-Floating Base: 6-DOF floating base joint for realistic dynamics
  • IMU Simulation: Pelvis-mounted IMU providing angular velocity, linear velocity, and orientation

​
Use Cases

​
1. Reinforcement Learning

Train locomotion policies using GPU-accelerated parallel simulation environments. The MuJoCo model integrates seamlessly with popular RL frameworks:
  • Isaac Lab / Isaac Gym
  • MuJoCo MJX (JAX-based)
  • Brax
  • dm_control

​
2. Controller Development

Test and tune control algorithms in simulation before hardware deployment:
  • PD controller gains
  • Model Predictive Control (MPC)
  • Whole-body control
  • Trajectory optimization

​
3. Gait Analysis

Analyze walking gaits, stability margins, and ground reaction forces:
  • Zero Moment Point (ZMP) analysis
  • Center of Mass (CoM) trajectory planning
  • Phase portrait analysis
  • Energy efficiency optimization

​
4. Sim-to-Real Transfer

Develop policies that transfer from simulation to the physical robot:
  • Domain randomization
  • System identification
  • Dynamics parameter tuning

​
Getting Started with MuJoCo

1

Install MuJoCo

Install MuJoCo 3.0+ and Python bindings:
pip install mujoco
2

Load the Model

Load the Asimov model from XML:
import mujoco

# Load the model
model = mujoco.MjModel.from_xml_path('sim-model/xmls/asimov.xml')
data = mujoco.MjData(model)

# Step the simulation
mujoco.mj_step(model, data)
3

Visualize

Launch the interactive viewer:
import mujoco.viewer

viewer = mujoco.viewer.launch(model)
4

Read Sensor Data

Access IMU and joint state data:
# Joint positions and velocities
qpos = data.qpos  # Joint positions
qvel = data.qvel  # Joint velocities

# IMU data
gyro = data.sensordata[0:3]      # Angular velocity
accel = data.sensordata[3:6]     # Linear acceleration
quat = data.sensordata[6:10]     # Orientation quaternion

​
Model Structure

The simulation model is organized into several key components: 
sim-model/
├── xmls/
│   └── asimov.xml          # Main MuJoCo model definition
└── assets/
    └── meshes/             # STL mesh files for visualization

​
Model Hierarchy

pelvis_link (floating base)
├── left_hip_pitch_link
│   └── left_hip_roll_link
│       └── left_hip_yaw_link
│           └── left_knee_link
│               └── left_ankle_pitch_link
│                   └── left_ankle_roll_link
│                       └── left_toe_link
└── right_hip_pitch_link
    └── right_hip_roll_link
        └── right_hip_yaw_link
            └── right_knee_link
                └── right_ankle_pitch_link
                    └── right_ankle_roll_link
                        └── right_toe_link

​
Next Steps

For detailed information about the MuJoCo model structure, joint configurations, and XML specifications, see:

MuJoCo Model Reference

Detailed breakdown of the asimov.xml model file, including joints, actuators, sensors, and collision geometry

​
Resources

Build docs developers (and LLMs) love

Get started for freeTalk to us