Quadruped_State_Estimation

Introduction

This repository includes the state estimator and the terrain estimator for the quadruped. The state estimator using Linear Kalman to estimate the position and velocity of the quadruped centroid. The terrain estimator estimates the normal vector of the quadruped located ground. The algorithm is validated on real robot Aliengo and can be deployed to other quadruped robots.

Installation

The project depends on Pinocchio, Eigen, ROS and runs on ubuntu2004.

ROS

You can follow the tutorial to install ROS Noetic.

Eigen

The version we use is Eigen 3.3.9. You can download the source code from the official website and install it.

Pinocchio

Pinocchio can be installed from tutorial.

Our project

mkdir -p catkin_ws/src
catkin_init_workspace
git clone https://github.com/skywoodsz/Quadruped-State-Estimation.git
cd ..
catkin_make

Run

1. Establish communication with the quadruped.

cd catkin_ws
sudo su
source devel/setup.bash
rosrun aliengo_control aliengo_low_level_node

2. Run the state estimator and the terrain estimator.

mon launch kf_estimator LegSlam_bringup.launch 

3. Visualize with RVIZ

mon launch kf_estimator rviz.launch

Framework

• aliengo_control: communicate with aliengo.
• cheetah_msgs: custom msg.
• unitree_description: Robot urdf model, used for forward kinematics.
• kf_estimator/state_estimate: kalman eatimator.
• kf_estimator/terrain_estimator: terrain_estimator.

ROS Topic

• /dog/imu_data: imu raw data.
• /dog/motor_data: quadruped joint data, including joint angles and velocities.
• /dog/leg_contact: quadruped contact Leg Information.
• /odom: state estimation result.
• /dog/terrain_norm: ground estimated normal vector.