Inertial SLAM Tutorial

This repo contains a basic inertial SLAM example using the implementation of the EKF from OpenVINS. This code is meant for tutorial purposes, to demonstrate how to set up an EKF for a basic SLAM problem where a robot moves in 3D space, collecting noisy inertial data and measurements to unknown landmarks in the body frame.

Dependencies

This code has the following dependencies

• CMake >= 3.10
• Eigen3 (>= 3.3)
• Boost
• glog
• yaml-cpp

Installation

This code has been tested on Ubuntu 20.04 and 22.04. The required dependencies can be installed using

sudo apt update
sudo apt install cmake libeigen3-dev libboost-all-dev libgoogle-glog-dev libyaml-cpp-dev

To build, clone the repo and run CMake

cd inertial_slam_tutorial
mkdir build && cd build
cmake ..
make -j4

Running the example

The main C++ example simulates a robot moving through 3D space along with noisy IMU and landmark measurements, and outputs state estimates and covariances. The example can either be run standalone or through the provided Python script, which will also evaluate and plot the results. To run the example through Python, first install the required Python dependencies and run the script as

pip install numpy matplotlib scipy
cd examples
python run_ekf_slam_example.py

References

This repo incorporates code from OpenVINS, where the EKF implementation is and B-Spline simulator are derived from.

TODO

• Add derivation of process and measurement model Jacobians