Geek Answers Handbook

Possible Datasets for Testing Path Finding Algorithms

I am working on finding ways.

So far, I have tested my code on scenes consisting of 2D cells. I also created a simple three-dimensional scene to test my work.

I would like to test my work on some 3D scenes ... but it takes a lot of time to create them.

Does anyone know of any scene datasets that I could use to test my path search algorithms?

+4
source share
2 answers

To get a better answer, you really need to specify the dimension of the configuration spaces you want to consider. You will not deal with the problems of protein folding and docking (more than 200 degrees of freedom) with discrete graph searches. Even relatively small planning problems (in terms of academic problems), about 6 degrees of freedom, can quickly become insoluble.

Most of the best planning examples are usually published in scientific articles and then make their way to wider use. Some of the best works are usually published in IEEE journals or at conferences Intelligent Robots and Systems (IROS) and the International Conference on Robotics and Automation (ICRA). It may also be useful to use a bibliography of a well-known reference in a given field, for example LaValle's “Traffic Planning” as a starting point for further research (available in bibtex here )

Mark Overmars work in computational geometry, and the planning communities made some of the issues discussed in his publications very recognizable. It is worth checking if his current students and colleagues of the graph have any data sets available at the moment.

If you are happy to still do some work in 2d and manually convert the image to geometric data, the Kris Beevers website has a number of processed examples for a number of planners in 2 workspaces.

The library of motion strategies contains a number of classical tasks of motion planning for use in workspaces 2d and 3d with different dimensions of the configuration space depending on the problem. It includes:

  • L Birdcage sections
  • trailers
  • several trailers
  • mazes
  • kinematic chains
  • nonholonomic cars

A later implementation of the motion planning training library is the open engine library developed by Kavraki . Due to licensing, I have not personally tested it, but I assume that they submit several examples and tests with their project.

Currently, within the framework of the OpenRAVE project, several significantly more complex examples of motion planning in kinodynamic mode are publicly available. Their gallery opens with an eye.

+1
source

when I need large 3D datasets, I usually use attractors or other time series. You just need to repeat as many times as you want, and it will create a good set of 3D data.

Try this "Peter de Jong Attractor":

Xn+1 = sin(a Yn) - cos(b Xn) Yn+1 = sin(c Xn) - cos(d Yn)

Where (for example): a = 1.4, b = -2.3, c = 2.4, d = -2.1

0
source

More articles:


All Articles