#14 The Hidden Complexity of Procedural Agent Spawning

The Problem

Spawning AI agents procedurally in multiplayer games involves far more than random placement. You need to coordinate timing across clients, validate positions against navigation meshes, persist state across sessions, and handle network authority - all while maintaining performance.

Core Challenges

1. NavMesh Timing

NavMesh isn't immediately available on game start. Solution: Hook into NavMesh.onPreUpdate and only collect positions when the navigation system is ready.

2. Network Coordination

Clients can't spawn directly - they must collect valid positions and send them to the server. Uses [SyncVar] to track spawn state across all clients.

3. Position Validation Pipeline

1. Generate random point within BoxCollider bounds
2. Snap to nearest NavMesh edge using FindClosestEdge()
3. Verify snapped position is still within bounds
4. Convert to global coordinates for persistence

This prevents ceiling spawns and ensures navigable positions.

4. Persistence

Agents spawn only once per location. The system:

• Tracks spawn state in save files using unique location IDs
• Checks LocationState.didSpawnedAgents on server start
• Integrates with entity system for unique IDs and coordinate conversion

5. Group Organization

Hierarchical randomization across four levels:

• Package types (civilians, guards, etc.)
• Groups per package (min/max counts)
• Agents per group (min/max counts)
• Agent types within groups (random selection from prefab arrays)

Positions are consumed as used to prevent overlapping spawns.

Performance Optimizations

Statistical Oversampling

Generates 3x more positions than needed since many fail NavMesh validation. Trades memory for reliability instead of expensive retry loops.

Coordinate Abstraction

Handles three coordinate systems: Unity local, global world (for persistence), and NavMesh. Enables worlds larger than Unity's native precision.

Key Insights

Timing: Event-driven architecture prevents race conditions when systems aren't ready.

State Management: Simple spawning becomes complex when adding persistence and networking.

Validation Layers: Procedural generation requires multiple fallback strategies for reliability.

Performance Trade-offs: Statistical approaches often outperform deterministic validation.

Conclusion

Procedural agent spawning in multiplayer games requires orchestrating event-driven networking, spatial validation, persistent state, and performance optimization. What appears simple on the surface reveals significant architectural complexity when implemented properly.

The challenge isn't the individual components - it's making them work together reliably in a networked, persistent, real-time environment.