A* Pathfinding Breaks at Scale.
This Replaces It.
A* was never designed for real-time, multi-agent systems. It works — until it doesn't. StrataNav replaces per-agent pathfinding with field-based navigation infrastructure.
The Core Problem
Traditional pathfinding works like this: for each agent, compute a path. When the goal changes, compute again. When the environment changes, compute again. At 10 agents, this is fine. At 100, it stalls. At 1000, it becomes the bottleneck.
| Metric | A* | StrataNav |
|---|---|---|
| 100-agent game tick | 111.7ms | 7.06ms |
| Replanning cost | Full re-search per agent | Near-zero (field lookup) |
| Multi-agent handling | Per-agent collision avoidance | Coordinated field movement |
| Frame budget impact | High (runs in-engine) | Zero (runs server-side) |
| Goal change cost | Recompute from scratch | Instant query update |
| Engine dependency | Tightly coupled | HTTP JSON — any engine |
The Architectural Difference
A* — Computation Model
Pathfinding is per agent, recomputed every time goals or environments change. Cost scales linearly (or worse) with agent count. Runs inside the game engine, consuming frame budget.
each agent: compute path → expensive goal changes → recompute obstacle moves → recompute repeat every tick
StrataNav — Infrastructure Model
Navigation is precomputed as a persistent field. All agents query the same structure. Goal changes require no recomputation. Runs server-side — zero frame budget consumed.
precompute field → once agent 1: query → microseconds agent 2: query → microseconds agent N: query → microseconds goal changes → just query again
Why A* Fails at Scale
A* is optimal for single-agent, static-environment pathfinding. It was designed in 1968 for exactly that use case. Modern games violate every assumption A* makes:
Hundreds of agents — each running independent graph searches.
Constant goal changes — every change triggers full recomputation.
Dynamic environments — obstacle movement invalidates cached paths.
The industry accepted this as a constraint to design around. StrataNav removes it.
When You Should Move Beyond A*
If you are doing any of the following, you are already at the limit:
→ Optimizing A* heuristics to squeeze out milliseconds
→ Batching A* queries to reduce per-frame cost
→ Capping agent count to stay within frame budget
→ Spending engineering time on pathfinding instead of gameplay
These are symptoms of an architectural limitation, not an algorithm problem. The fix is not better A* — it is replacing A* with infrastructure.