0.9.0

Features

• Add Nesterov-accelerated GJK

Bugfixes

• Fix numerical problem in Jolt's GJK implementation.
• Fix numerical problem in MPR.
• Use time.perf_counter() for benchmark.

0.8.0

Features

• Fast AABB tree implementation for broad phase, accelerated with numba #61
• Hydroelastic contact model supports Young's modulus as input #62
• Script to compute potentials of tetrahedral mesh #56
• Add hydroelastic BVH for loading URDFs #68

0.7.1

Bug Fixes

• Update to API of pytransform3d 2.0.0

0.7.0

Features

• Improvements of hydroelastic contacts
  • Faster tetrahedron intersections with AABB checks
  • Tetrahedral box meshes #50
  • Tetrahedral ellipsoid meshes #51
  • Tetrahedral cylinder meshes #52
  • Tetrahedral capsule meshes #53

0.6.0

Features

• Ellipse collider #34
• GJK from Jolt Physics #38
  • GJK intersection test is about 30-40% faster
  • GJK distance calculation is about 3-5x faster
• Adds distance3d.geometry.barycentric_coordinates_tetrahedron
• Adds hydroelastic contact model (module: distance3d.hydroelastic_contact) #44
• Adds function to load tetrahedral meshes: distance3d.io.load_tetrahedral_mesh

0.5.0

Features

• Use trimesh as a fallback solution to load meshes from URDF files when Open3D does not work
• Collision margin collider #30
• Containment tests #32

Performance

• Faster inversion of transformation matrices with numba

0.4.0

Features

• Add support function for box and sphere
• Cone, disk, and ellipsoid colliders #24 #14
• Hill climbing to speed up support function of convex meshes #15
• Minkowski portal refinement (MPR) for collision detection #18, about 20x faster for collision detection than GJK
• Fast GJK intersection test #26, 20x faster than standard GJK and 10% faster than MPR

Performance

• Acceleration with numba #13 , for example:
  • point_to_triangle is 7x faster
  • point_to_plane is 2.5x faster
  • point_to_disk is 6x faster
  • line_to_line_segment is 6x faster
  • line_segment_to_line_segment is 5x faster
  • line_segment_to_plane is 5x faster
  • line_to_triangle is 15x faster
  • line_segment_to_triangle is 14x faster

Refactoring

• Refactored GJK for readability and minor speed improvements #11
• Refactored colliders #23
• All support functions have name prefix support_function_
• Move BoundingVolumeHierarchy to new module broad_phase
• Rename Convex to ConvexHullVertices

0.3.0

Features

• Distance
  • from line segment to circle
  • from plane to plane
  • from plane to triangle
  • from plane to rectangle
  • from plane to box
  • from plane to ellipsoid

Performance Improvements

• Significantly faster AABB computation for boxes, cylinders, and capsules
• Significantly faster self collision detection by pruning
• 6x faster distance from point to box

Full Changelog: 0.2.0...0.3.0

0.2.0

New Features

• Self collision detection for robots
• Distance computation
  • from line to plane
  • from line segment to plane
  • from line to circle
  • from disk to disk

0.1.1

Fix image URL for PyPI