Slip-Robust InEKF for Wheeled Robots

This project aims to estimate the states of a wheel robot in outdoor environments using IMU data, wheel encoder data. An InEKF based approach is taken to achieve the objective. This project is developed based on the "Legged Robot State Estimation in Slippery Environments Using Invariant Extended Kalman Filter with Velocity Update" and the open source code of "In-EKF".

Getting Started

To run the Python code, the following environment and dependencies are required:

• Python == 3.6
• numpy == 1.20.3
• matplotlib == 3.4.3
• scipy == 1.7.1
• pandas == 1.3.4

Code organization

.
├── results                     # Folder contains final results images
├── src                         # Matlab and Python scripts
│   ├── Matlab                  # Matlab scripts
│   │   ├── main.m              # Main script of state estimation with In-EKF in indoor environment
│   ├── Python                  # Python scripts
│   │   ├── helper_func.py          # Helper function
│   │   ├── husky_inekf_indoor.py   # Main script of state estimation with In-EKF in indoor environment
│   │   ├── husky_inekf_light.py    # Main script of state estimation with In-EKF in parking lot environment
│   │   ├── inekf_imu_cameraPos.py	# In-EKF with encoder and camera measurements
└── README.md

Running the code

To run the Python code, run husky_inekf_light.pyfor parking lot environment

To run the Matlab code, run main.mfor parking lot environment

Implementations

• See the [report] (upcoming soon!) for detailed implementations.

Results

The following video shows how our proposed algorithms (encoder-only and encoder-pseudo measurement) work well compared to ORB-SLAM2, Visual Odometry and GPS.

Contributors

• Theodor Chakhachiro: Github, Linkedin, Google Scholar, Homepage

• Ryan Feng: Linkedin, Homepage

• Vladimir Krokhmal: Github, Linkedin

• Xihang Yu: Github, Linkedin

• Hongyu Zhou: Github, Linkedin, Google Scholar, Homepage