Agent Architecture

Overview

The core differentiation of SynCity lies in the individual intelligence of its citizens. Each agent is not merely a sprite on a screen but a persistent data object with an internal state machine. The Agent Architecture defines how these entities process information from their environment and execute actions.

The Decision Loop

Every agent operates on a cyclic logic loop known as a "Tick." During each simulation tick, the agent performs three distinct phases.

1. Perception (Input) The agent scans its immediate surroundings and internal status.

• Internal Data: Current energy, hunger, bank balance, and employment status.

• External Data: Distance to workplace, price of nearby goods, and available housing.

2. Evaluation (Logic) The agent uses a Utility-Based Scoring System to rank potential actions. The system assigns a score to every possible action based on the agent's current needs.

• Example: If "Hunger" is high, the action "Buy Food" receives a score of 90/100. If "Energy" is high, the action "Sleep" receives a score of 10/100.

3. Execution (Output) The agent selects the action with the highest score and executes it. This might involve pathfinding to a location, transferring credits, or updating a database record.

Priority Ranking System

To prevent chaotic behavior, the AI follows a strict hierarchy of priorities, similar to a simplified version of human psychology.

1. Critical Survival: Actions required to prevent agent "death" or removal from the city (e.g., eating, sleeping).

2. Economic Security: Actions required to maintain solvency (e.g., going to work, paying rent).

3. Self-Improvement: Actions that increase long-term value (e.g., studying, learning a new skill).

4. Leisure: Actions that increase happiness (e.g., visiting a park, shopping).

State Persistence

Agents retain memory of their actions. If an agent applies for a job and is rejected, they will "remember" this cooldown period and will not apply to the same company immediately. This reduces computational load and mimics realistic human hesitation.