khshmt / ekf_slam

A Sparse Line-Based EKF SLAM System for Indoor Environment

Sparse Line-Based EKF SLAM System for Indoor Environment

1. Introduction:

This project implemented an EKF SLAM system for indoor differential robot, which use laser range finder as its main sensor. Assume no prior knowledge of environment, can simultaneously build a map of sparse lines and localize robot, no limitation of map size.

2. V-REP Simulation Environment:

This project V-REP 3.5.0 as the simulation environment, the scene file is provided in scene/scene.ttt

3. Prerequisites

Built and tested on Ubuntu 16.04

Dependencies:

1. ROS Kinetic
2. Eigen: 3.3.5

4. Run Example:

1. Launch roscore

roscore

2. Open V-REP software and open the provided scene file /scene/scene.ttt
3. launch EKF SLAM system

roslaunch ekf_slam ekf_slam.launch

Then use 'W' 'A' 'S' 'D' to control robot to move up, left, down and right

5. Mathematical Derivation

The complete mathematical derivation of whole system can be found in Mathematical Derivation.pdf file

6. Result:

The white line is ground truth of robot trajectory, and the yellow line is posterior estimation of robot trajectory.

Figure 1 Normal trajectory

Figure 2 Trajectory with unexpected collision

The trajectory contains four unexpected collision. When collision occurs, the posterior estimation will not be as accurate as normal, but the system can recover from collision error afterwards, which shows the robustness of the system