Non-convex polygons clustering algorithm

12
Non-convex polygons clustering algorithm Ekaterinburg 2016 Presenter: Mikhail Vasilenko

Transcript of Non-convex polygons clustering algorithm

Page 1: Non-convex polygons clustering algorithm

Non-convex polygons clustering algorithm

Ekaterinburg 2016

Presenter: Mikhail Vasilenko

Page 2: Non-convex polygons clustering algorithm

Introduction

• Use all obstacles

• How to decrease calculation time?

• How to optimize for quick motion planning?

Page 3: Non-convex polygons clustering algorithm

Approach

• Rectangles (frames)

• Hierarchy

• Different properties on different levels

Page 4: Non-convex polygons clustering algorithm

Frames to Containers

Inclusion? Intersection? One-element container?

Page 5: Non-convex polygons clustering algorithm

Containers to Family

Intersection? Chains of containers? One-element family?

Page 6: Non-convex polygons clustering algorithm

Families to Groups

Inclusion? Intersection? Close distance? One-element group?

Page 7: Non-convex polygons clustering algorithm

Step-by-step solution

Page 8: Non-convex polygons clustering algorithm

Parallel approach

Page 9: Non-convex polygons clustering algorithm

Summary• Described algorithm is implemented in C++

• Tested on real-world data

• Algorithm is used in route-planning project

• Integration of parallelism is underway

Page 10: Non-convex polygons clustering algorithm

Thanks for your attention!

Page 11: Non-convex polygons clustering algorithm

• To cluster families into groups each of the former is assigned a rough hull

• All the groups have non-intersecting rough hulls

• Merges groups with distances lower than twice that value

Last step

Page 12: Non-convex polygons clustering algorithm

Test data

• The case depicted had 1031 polygons

• which were clustered into 310 containers

• 201 families

• and 12 families groups.