Cloth Simulation

Implementation of Cloth Simulation Using the Mass-Spring and Verlet Models in UE4

Period: September 1, 2022 → December 6, 2022
Last Edited: February 25, 2025, 3:05 PM
Type: Game, Portfolio
Author: MINHA JEON
Tools & Language: Blender, C++, Unreal
State: Done
Description: Implementation of Cloth Simulation using the Mass-Spring and Verlet models in UE4
Participants: 3

Kyung Hee University 2022–2 Game Contents Capstone Design

1️⃣ Overview

  • Goal: Develop a 1 vs. 1 fighting game incorporating a Cloth Simulation based on a Mass-Spring model
  • Purpose: Enhance the game’s sense of realism and dynamism by introducing realistic cloth behavior
  • Key Points
    • Study and develop cloth simulation technology that can be applied to a game
    • Implement a 1 vs. 1 fighting game with ranged and melee attacks, environment interactions, and distinct animations triggered by different hit locations
  • Tools: Visual Studio 2017, UE4 (4.26)
  • Role: Implement and optimize the Cloth Simulation, integrate it into the game, troubleshoot issues, and suggest directions for improvement

2️⃣ Survey & Development Process

> Technology

  1. Cloth Simulation
  • Mass-Spring Model
    • Each vertex of the cloth is treated as a mass (particle), and springs form the connections (edges) between vertices.
    • Edge=spring, Vertex=mass
    • Use Hook’s law, Newton’s 2nd Law as an equation of motion.
    • Particles are connected by several types of spring, which results in a spring force.
    • In addition, external force (gravity, wind, collisions with body..) acts to achieve final cloth simulation.
  • Basic Cloth Simulation Steps
    • Calculating forces (spring, damping, gravity, etc.)
    • Integrating positions over time
    • Handling collisions and constraints

  1. Our Approach

    • Used a Mass-Spring model combined with implicit time integration (Backward Euler) for Cloth Simulation
    • Created and updated the cloth mesh via procedural mesh in UE4
    • Integrated the Eigen library into UE4 to improve computational speed
    • Began real-time optimization afterward

  2. Issues Encountered

    • ⛔ Issue 1: Time Solve & Fix Points
 - Encountered errors during the time-solve phase and when fixing anchor points.
 - Due to structural differences, null values appeared when using Eigen for computation, preventing proper cloth updates.
 - Solved by adjusting the relevant data structures.
  • ⛔ Issue 2: Cloth Update
 - The cloth converged and did not update correctly at runtime.  
 - Suspected a mismatch between the Mass-Spring particle indices and the vertex indexing in the procedural mesh.  
 - Forces were not mapped correctly, causing the cloth to either contract or deform abnormally.  
 - Time constraints made it difficult to resolve fully, so we decided to pivot to a different integration approach.
  1. Pivot to PBD & Verlet
  • Position Based Dynamics (PBD)
    • Performs physics calculations based on positions rather than velocities, simplifying certain steps.
    • Uses a system of constraints directly on particle positions.
  • Verlet Integration
    • An explicit approach that operates on positions alone.
    • No need to store or compute velocity for each particle; updates rely on previous and current positions.
    • Constraints are solved by projecting positions directly, rather than by applying forces.

> Game

  • 1 vs. 1 Fighting Game in an Underground Arena
  • Implements an AI opponent that reacts to player actions
  • Dodge/Combo system

  • Distinct reactions and hit stops depending on which body part is struck

  • Emphasizing Cloth Simulation
    • Added banners, flags, and other cloth items to showcase realistic cloth movements
    • Cloth items (like towels) scattered around the ring; can be picked up and thrown at the opponent
  </div>

</div>

3️⃣ Result

🔗YouTube Video

Level of Achievement

Significance & Limitations

  • Significance

    • Cloth Simulation can apply to numerous in-game elements (character clothing, background elements)
    • Allows physically driven, more realistic motion compared to purely pre-animated effects

  • Limitations

    • Did not implement advanced deformation (e.g., damage or tearing) of cloth during actual combat
    • Real-time CPU simulation proved demanding; further performance gains may require GPGPU (GPU-based) solutions

🔗 Related Resources & Links

GitHub
https://github.com/NCTp/GameCapstoneDesignProject

YouTube
[2022-2 GCCD] Team 3 Cloth Simulation

Reference