Choosing and Evaluating 2D Physics Engines for Mechanics Experiments: Issues and Solutions

We selected an open‑source 2D physics engine for developing mechanics experiments and encountered several challenges during selection and early development.

Engine selection : Although Box2D is widely regarded as the best 2D physics engine, the Nape engine (https://github.com/deltaluca/nape) offers 3‑6× higher performance and half the memory usage compared to the Flash version of Box2D. Nape’s API is more JavaScript‑friendly, while Box2D’s API, originally written in C++, differs significantly from JavaScript conventions.

Test scenarios demonstrated differences between the engines:

Gravity simulation (9.8 m/s²) with 1 m = 5000 px and 1 cm objects showed better convergence in Nape.

Spring force gauge required both a spring joint and a line joint; Nape provides LineJoint, Box2D does not.

Pulley systems: Nape supports four‑body pulleys, enabling movable pulleys, whereas Box2D limits pulleys to two bodies.

Buoyancy: Box2D’s current version lacks built‑in buoyancy; Nape can implement it via gravMass.

Friction: Nape distinguishes static, dynamic, and rolling friction, while Box2D offers only a single friction coefficient.

Concave polygons: Nape includes APIs for concave polygon decomposition, which Box2D lacks.

Dynamic constraint addition: Nape allows toggling constraint activity without frequent creation/destruction, reducing garbage collection overhead.

Hooks and light ropes: Nape permits separate settings for gravMass (zero) and actual mass, solving contradictory mass requirements.

Despite choosing the higher‑performance engine, precision remained insufficient. The article identifies four main causes: (1) scaling of gravity and object size, (2) multiple constraints acting simultaneously leading to non‑convergence, (3) data‑precision limitations exposing oscillations, and (4) inherent oscillation tendencies.

Additional issues include mismatched mass‑density‑volume relationships in 2D simulations, friction modeling discrepancies, and complex buoyancy implementation.

Proposed solutions :

Increase engine frame rate or reduce time step to improve accuracy without altering gravity.

Extend or customize constraints where feasible.

Adopt a 3D physics engine to handle realistic mass‑density‑volume calculations.

Set Nape’s staticFrictionThreshold to zero to address friction and sliding problems.

Use gravMass or add counter‑impulses to compensate for buoyancy‑induced mass changes.

References: [1] Nape Physics Engine (http://hanyeah.com/mirror/nape/help/manual.html), [2] Box2D documentation (https://box2d.org/documentation/).