Position-Based Dynamics (PBD)

The core principle of PBD is its direct position-based solver. Unlike force-based methods (like penalty forces or impulse-based solvers) that apply corrective forces and rely on integration, PBD directly calculates and projects particle positions to satisfy constraints. This involves:

• Defining a set of constraint functions (e.g., distance constraints for cloth, volume constraints for fluids).
• Iteratively solving these constraints by projecting each particle's position to the nearest valid state.
• The result is a highly stable simulation that is largely independent of time step size, avoiding the explosive instabilities common in stiff spring systems.

PBD provides unconditional stability for any time step, a critical feature for real-time applications like games and virtual reality. This is achieved because:

• The solver works directly on positions, bypassing the explicit integration of velocities that can blow up with large time steps.
• Constraints are solved via a Gauss-Seidel-type iteration, where each constraint is solved locally and sequentially. This iterative projection naturally damps energy, preventing runaway oscillations.
• While this damping introduces numerical dissipation (loss of energy), it is a controlled trade-off for guaranteed stability, making it ideal for interactive simulations where "not crashing" is more important than perfect energy conservation.

PBD models objects entirely through geometric constraints rather than material properties. Common constraints include:

• Distance Constraints: Maintain a fixed length between two particles (modeling cloth or mesh edges).
• Bending Constraints: Resist bending between adjacent triangles in a mesh.
• Volume Constraints: Preserve the total volume of a closed mesh (for soft bodies).
• Collision Constraints: Enforce non-penetration with obstacles or other bodies.
• Shape Matching Constraints: Guide a deformed cluster of particles back to its original shape (for elastic objects). This formulation is highly modular; new materials and behaviors can be modeled by simply designing new constraint functions.

PBD is engineered for real-time performance, often achieving millisecond-level computation times per frame. This is enabled by:

• Linear Computational Complexity: The per-constraint calculations are simple vector operations (additions, multiplications, normalizations).
• Controlled Iteration Count: The number of solver iterations is fixed (typically 3-10), providing a predictable performance budget. More iterations increase accuracy but also cost.
• Mass-Splitting: During constraint projection, position corrections are weighted by inverse mass, allowing for static anchors or objects of different weights with minimal overhead.
• Parallelizability: The local nature of most constraints makes PBD highly amenable to parallel processing on GPUs, which is leveraged by modern engines like NVIDIA PhysX.

Due to its stability and speed, PBD is the de facto standard for several simulation domains in interactive media:

• Cloth and Fabric Simulation: Used extensively in games and film for character clothing, flags, and banners.
• Hair and Fur: Models strands as chains of particles connected by distance and bending constraints.
• Soft Body Objects: Simulates squishy objects, inflatables, and character flesh using shape-matching and volume constraints.
• Fluid and Granular Materials: While not as accurate as SPH or MPM, PBD can approximate fluids and sand using density constraints (as in the original "Position Based Fluids" paper).
• Interactive Deformables: Ideal for VR environments where users can directly poke, pull, and tear deformable objects in real-time.

PBD differs fundamentally from traditional physics simulation approaches:

• Vs. Force-Based (Penalty Methods): Penalty methods add strong spring forces to enforce constraints. These create stiff differential equations requiring very small time steps for stability. PBD is stable at large time steps.
• Vs. Impulse-Based Methods: Impulse-based solvers (common in rigid body engines) resolve collisions by instantaneously changing velocities. PBD resolves collisions by directly correcting positions, which can be more robust for persistent contact and stacking.
• Vs. Finite Element Method (FEM): FEM provides high physical accuracy for stress and strain but is computationally expensive and complex. PBD sacrifices material accuracy for robustness, speed, and artistic controllability, making it a geometric rather than a continuum mechanics approach.

PBD is the dominant method for real-time cloth simulation in games and interactive media. It models fabric using a network of particles connected by distance constraints (for stretching) and bending constraints (for folding).

• Key Advantage: Unconditional stability prevents explosive behavior common in force-based methods, even with large time steps.
• Typical Constraints: Stretching, shearing, bending, and self-collision.
• Example: Simulating dynamic character clothing, flags, and banners in video games.

PBD efficiently simulates soft, squishy materials like rubber, jelly, flesh, and inflatable objects. Objects are typically represented by a volumetric mesh or a set of particles with shape-matching constraints that pull the deformed shape back towards its original rest state.

• Shape Matching: A core PBD constraint that groups particles and applies a best-fit transformation to maintain global shape.
• Applications: Character blobs, interactive game objects, and basic biological tissue modeling.
• Benefit: Provides visually plausible deformation without solving complex finite element equations.

While not as accurate as SPH for scientific simulation, PBD is used for interactive fluid visualization. Fluid particles are governed by density constraints to maintain incompressibility and cohesion constraints for surface tension.

• PBD vs. SPH: PBD's constraint-based approach can be more stable and faster for small-scale, visually-focused effects.
• Common Use Cases: Splashing water, magical effects, and viscous fluids in real-time applications.
• Limitation: Often trades physical accuracy for visual appeal and performance.

Strands of hair or fur are simulated as chains of particles connected by stiff distance constraints. Bending and twisting constraints add realism, while collision constraints prevent hair from clipping through the head and body.

• Performance: PBD handles the hundreds of individual strands required for a full head of hair in real-time.
• Control: Artists can easily tune stiffness and damping parameters to achieve the desired look.
• Industry Standard: Widely used in game engines and digital content creation tools for character finaling.

PBD can simulate rigid bodies by treating them as particles with shape-matching constraints. For destruction, a pre-fractured object's pieces are connected by breakable constraints. When a force threshold is exceeded, constraints break, simulating fracture.

• Advantage: The unified solver handles mixed scenarios (e.g., cloth on a breaking table) seamlessly.
• Use Case: Destructible environments in games where predictable, performant breaking is needed.
• Method: Often combined with a traditional rigid body solver for primary motion, using PBD for secondary fracture effects.

In robotics, PBD's constraint-based framework is being explored for motion planning and controller design. It allows for the direct specification of task-space constraints (e.g., "end effector must reach here") which are solved efficiently.

• Inverse Kinematics (IK): PBD can solve IK problems by treating target positions as constraints.
• Physical Interaction: Simulating compliant contact and manipulation of soft objects in a robot's environment.
• Research Area: Extending PBD to model more accurate friction and motor dynamics for sim-to-real transfer.

A numerical technique for simulating deformable bodies by discretizing a continuous object into a mesh of small elements and solving for stress and strain. Unlike PBD's direct position manipulation, FEM solves partial differential equations to compute deformation forces.

• Primary Use: High-accuracy simulation of materials under stress (e.g., structural analysis, crash testing).
• Key Difference: FEM is typically computationally expensive and used for offline engineering, whereas PBD prioritizes real-time performance for interactive applications.

A mesh-free, Lagrangian computational method used to simulate fluid flows and other continuum media using interacting particles. SPH and PBD are both particle-based but differ in their governing equations.

• Primary Use: Real-time and offline fluid simulation (e.g., water, smoke, lava).
• Key Difference: SPH solves the Navier-Stokes equations for fluid dynamics, while PBD uses geometric constraints to enforce incompressibility and other properties.

A hybrid Eulerian-Lagrangian technique that combines particle-based representation with a background grid. It excels at simulating complex materials with extreme deformations and phase changes.

• Primary Use: Simulating snow, sand, mud, and viscoelastic fluids with cutting and fracturing.
• Key Difference: MPM accurately handles history-dependent materials and plastic deformation, whereas standard PBD is better suited for elastic, cloth-like behavior.

An algorithmic component of a physics engine that resolves forces and impulses to satisfy physical constraints, such as joints or contact non-penetration. PBD is itself a type of constraint solver that works directly on positions.

• Primary Use: Ensuring simulated objects behave according to defined physical rules (e.g., a door hinge, a ball resting on a floor).
• Key Relationship: Traditional force-based solvers (like the Projected Gauss-Seidel solver) compute forces to satisfy constraints; PBD skips the force calculation and projects positions directly.

A broad category of physics simulation that models objects capable of deformation, such as cloth, rubber, or flesh. PBD is a leading method for real-time soft body simulation.

• Primary Use: Character jiggle, clothing, inflatables, and organic-looking objects in games and films.
• Implementation Methods: Includes mass-spring systems, FEM, and Position-Based Dynamics. PBD is often favored for its numerical stability and ease of controlling stretching and bending stiffness.

An iterative numerical method used in traditional, force-based physics engines to solve constraint problems like contact and joints. It is the primary alternative solver architecture to PBD.

• Primary Use: Solving Linear Complementarity Problems (LCPs) for contact and friction in rigid body simulations.
• Key Difference: PGS solves for impulses and Lagrange multipliers in the velocity domain, while PBD performs positional projections. PBD often provides more controllable stiffness and avoids the overshooting issues that can occur with force-based solvers.