IP Library Granted Patent US 12,734,449
Granted Patent B2
US 12,734,449 · App. 18/908,916 · Granted Sep 15, 2026

Generating compound colliders

Inventor: Philip Cockram (London, GB)
Assignee: Sony Interactive Entertainment Inc.
A63F13/577A63F13/52A63F13/54A63F13/56A63F13/57G06T13/40G06T17/00A63F13/63A63F13/79G06T17/205G06T2207/30196
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,734,449
App. No.
18/908,916
Granted
Sep 15, 2026
Kind
B2
Abstract

A method of generating a compound collider for a given object, the method comprising the steps of: receiving a mesh of the given object, the mesh comprising at least one concave portion; and generating a compound collider representative of at least the concave portion of the mesh, in which the step of generating comprises: inputting the mesh to a machine learning model trained to generate a respective compound collider based on an input mesh comprising at least one concave portion, and using the trained machine learning model to generate the compound collider representative of at least the concave portion of the mesh.

Claims (40)

1 . A computer-implemented method of generating a compound collider for a given object, the method comprising:

receiving, by one or more processors, a mesh of the given object, the mesh comprising at least one concave portion;

receiving, by the one or more processors, information representative of a bone hierarchy of the given object;

generating, by the one or more processors, a compound collider representative of the at least one concave portion of the mesh, wherein the generating comprises:

inputting the mesh and the information representative of the bone hierarchy to a machine learning model,

wherein the machine learning model has been trained to learn a correspondence between an input mesh and an input representative of an input bone hierarchy to generate a respective compound collider configured to move with the input bone hierarchy; and

using the trained machine learning model to generate the compound collider based on a learned correspondence between the mesh of the given object and the information representative of the bone hierarchy of the given object, the generated compound collider being configured to move with the bone hierarchy of the given object.

2 . The method of claim 1 , wherein the machine learning model is trained to generate the compound collider by associating at least one respective convex collider of the compound collider with a respective bone of the bone hierarchy of the given object.

3 . The method of claim 1 , wherein the information representative of the bone hierarchy includes information representative of a range of motion of respective joints comprised by the bone hierarchy.

4 . The method of claim 1 , wherein the information representative of the bone hierarchy includes information representative of one or more animations for the given object.

5 . The method of claim 4 , wherein the machine learning model is trained to generate, based on the mesh, the bone hierarchy, and the one or more animations, a respective compound collider optimized for the one or more animations for the given object.

6 . The method of claim 1 , wherein the machine learning model is configured to generate the compound collider by generating a plurality of convex colliders in an order based on a hierarchical structure of the bone hierarchy.

7 . The method of claim 6 , wherein the machine learning model generates a first convex collider corresponding to a root bone of the bone hierarchy before generating a second convex collider corresponding to a child bone of the root bone.

8 . The method of claim 1 , wherein the machine learning model is trained to generate the compound collider from a plurality of predefined convex colliders.

9 . The method of claim 8 , wherein the plurality of predefined convex colliders comprises a plurality of primitive colliders comprising at least one of a cuboid, a cylinder, or a sphere.

10 . The method of claim 1 , wherein the machine learning model is trained to generate the compound collider based on a given category of object.

11 . The method of claim 10 , wherein the given category of object comprises at least one of a user avatar, an interactable object, or a non-player character (NPC).

12 . A system comprising:

one or more computer processors; and

one or more storage devices storing instructions that, when executed by the one or more computer processors, cause the one or more computer processors to perform operations for generating a compound collider for a given object, the operations comprising:

receiving a mesh of the given object, the mesh comprising at least one concave portion;

receiving information representative of a bone hierarchy of the given object;

generating a compound collider representative of the at least one concave portion of the mesh, wherein the generating comprises:

inputting the mesh and the information representative of the bone hierarchy to a machine learning model,

wherein the machine learning model has been trained to learn a correspondence between an input mesh and an input representative of an input bone hierarchy to generate a respective compound collider configured to move with the input bone hierarchy; and

using the trained machine learning model to generate the compound collider based on a learned correspondence between the mesh of the given object and the information representative of the bone hierarchy of the given object, the generated compound collider being configured to move with the bone hierarchy of the given object.

13 . The system of claim 12 , wherein the machine learning model is trained to generate the compound collider by associating at least one respective convex collider of the compound collider with a respective bone of the bone hierarchy of the given object.

14 . The system of claim 12 , wherein the information representative of the bone hierarchy includes information representative of a range of motion of respective joints comprised by the bone hierarchy.

15 . The system of claim 12 , wherein the information representative of the bone hierarchy includes information representative of one or more animations for the given object.

16 . The system of claim 15 , wherein the machine learning model is trained to generate, based on the mesh, the bone hierarchy, and the one or more animations, a respective compound collider optimized for the one or more animations for the given object.

17 . The system of claim 12 , wherein the machine learning model is configured to generate the compound collider by generating a plurality of convex colliders in an order based on a hierarchical structure of the bone hierarchy.

18 . The system of claim 17 , wherein the machine learning model generates a first convex collider corresponding to a root bone of the bone hierarchy before generating a second convex collider corresponding to a child bone of the root bone.

19 . The system of claim 12 , wherein the machine learning model is trained to generate the compound collider from a plurality of predefined convex colliders.

20 . One or more non-transitory computer-readable storage media storing instructions that, when executed by one or more computer processors, cause the one or more computer processors to perform operations for generating a compound collider for a given object, the operations comprising:

receiving a mesh of the given object, the mesh comprising at least one concave portion;

receiving information representative of a bone hierarchy of the given object;

generating a compound collider representative of the at least one concave portion of the mesh, wherein the generating comprises:

inputting the mesh and the information representative of the bone hierarchy to a machine learning model,

wherein the machine learning model has been trained to learn a correspondence between an input mesh and an input representative of an input bone hierarchy to generate a respective compound collider configured to move with the input bone hierarchy; and

using the trained machine learning model to generate the compound collider based on a learned correspondence between the mesh of the given object and the information representative of the bone hierarchy of the given object, the generated compound collider being configured to move with the bone hierarchy of the given object.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Oct 8, 2024
From: COCKRAM, PHILIP
To: SONY INTERACTIVE ENTERTAINMENT INC.
Reel/Frame 068825/0551 →
Priority Claims (1)
GB 2316062 · Oct 20, 2023 · national
Continuity (1)
Related Publication 20250128162A1 · Apr 24, 2025
References Cited (29)
US 10282885B2 · Angelidis · 2019 [cited by examiner]
US 10431000B2 · Zhang · 2019 [cited by examiner]
US 10672190B2 · McBeth · 2020 [cited by examiner]
US 11508167B2 · Deng · 2022 [cited by examiner]
US 11941739B1 · Radzihovsky · 2024 [cited by examiner]
US 11978268B2 · Deng · 2024 [cited by examiner]
US 12175368B2 · Teig · 2024 [cited by examiner]
US 12450812B2 · Wei · 2025 [cited by examiner]
US 12572219B1 · Voros · 2026 [cited by examiner]
US 20150325030A1 · Witkin · 2015 [cited by examiner]
US 20180189999A1 · Angelidis · 2018 [cited by examiner]
US 20190026942A1 · Zhang · 2019 [cited by examiner]
US 20190108681A1 · Mcbeth · 2019 [cited by examiner]
US 20210319209A1 · Deng · 2021 [cited by examiner]
US 20230078756A1 · Deng · 2023 [cited by examiner]
US 20230290032A1 · Wei · 2023 [cited by examiner]
US 20240009560A1 · Liu · 2024 [cited by examiner]
US 20250061639A1 · Sanocki · 2025 [cited by examiner]
US 20250095261A1 · Molla · 2025 [cited by examiner]
US 20250128162A1 · Cockram · 2025 [cited by examiner]
US 20250278880A1 · Green · 2025 [cited by examiner]
US 20250299444A1 · Ye · 2025 [cited by examiner]
US 20250308148A1 · Joudrey · 2025 [cited by examiner]
US 20250371786A1 · Troy · 2025 [cited by examiner]
Combined Search and Examination Report for corresponding GB Application No. 2316062.5, 5 pages, dated Apr. 3, 2024. [cited by applicant]
Extended European Search Report for corresponding EP Application No. 24204445.1, 8 pages, dated Feb. 28, 2025. [cited by applicant]
Zhiqin Chen et al: “BSP-Net: Generating Compact Meshes via Binary Space Partitioning”, arxiv.org, Cornell University Library, 201 Olin Library Cornell University Ithaca, NY 14853, 10 pages, dated Dec. 7, 2020. [cited by applicant]
Boyang Deng et al: “CvxNet: Learnable Convex Decomposition”, arxiv.org, Cornell University Library, 201 Olin Library Cornell University Ithaca, NY 14853, 14 pages, dated Sep. 12, 2019. [cited by applicant]
Anonymous: “How to make a custom collider that will deform with a skeleton in Godot—Stack Overflow”,,Aug. 27, 2021 (Aug. 27, 2021), pp. 1-2, XP093251342,Retrieved from the Internet: URL:https://stackoverflow.com/questio… [cited by applicant]