agv path smoothing - A generating/fitting specified NURBS curve ros package

This is a ROS package developed NURBS method for path smoothing in autonomous mobile robot navigation to avoid global planners generated path which has many sharp turns. The NURBS(Non-Uniform Rational B-Spline) curve, consist of control point, knot vector and weight vector, is parameter curve that can let curve curvature continuity and with high degree of freedom to modify curve.

The main features include:

• Implemented the generate Line, Bezier, B-spline and NURBS curve.
• Implemented the global interpolation method (see reference) to fit discrete waypoints.
• Implemented useful function for motion planning to use, such as calculation NURBS curve derivative, curvature, curve length and so on.
• Implemented limited fitting function let the mobile robot do not need to rotate in place when navigating.

The package is under development, some future works are listing below:

• Knot insert function
• Collision checking
• Real time control point repositioning (path modification)
• Extend to 3D NURBS curve.

Installation

This package based on Eigen library to development, so please install eigen library first.

sudo apt-get install libeigen3-dev

Next, clone this repository to your catkin workspace and catkin build this package.

cd ~/catkin_ws/src
glt clone https://github.com/q576333/agv_path_smoothing
catkin build

Usage

Gerenting NURBS curve example

Setting control point, knot vector and weight vector in launch file

<rosparam param="control_point" > [0, 0, 0, -2, 4, -2, 4, 0, 4, 2, 0, 2, 0, 0]</rosparam> 
<rosparam param="knot_vector" > [0, 0, 0, 0.25, 0.5, 0.5, 0.75, 1, 1, 1]</rosparam>
<rosparam param="weight_vector" > [1, 0.5, 0.5, 1, 0.5, 0.5, 1]</rosparam>

Running

roslaunch agv_path_smoothing spline_visualization.launch

Fitting waypoints to NURBS curve example

Setting discrete waypoints and weight vector in launch file

<rosparam param="fitting_point" > [-6, -3, -5, -2.5, 1, 3, 2, 1.5, 3, 1, 7, 1.5, 9, -2.5]</rosparam>
<rosparam param="weight_vector" > [1, 1, 1, 1, 1, 1, 1]</rosparam>

Running

roslaunch agv_path_smoothing fitting_test.launch

Reference

• "The NURBS Book" Les Piegl and Wayne Tiller, 2nd edn., Springer, 1997.