ADR-010: Decoupled Bot AI (Autopilot Strategy)

• Status: Accepted
• Date: 2026-01-06
• Deciders: Simon

Context and Problem Statement

Initially, bot logic (pathfinding, targeting, decision-making) might have been tightly coupled with the Bot or Tank class. This made the logic hard to reuse, difficult to test, and complicated the core entity code.

Decision Drivers

• Modularity: Separation of AI behavior from entity logic.
• Testability: Independent testing of pathfinding and decision-making.
• Extensibility: Easier to add or swap different AI strategies.

Considered Options

1. Embedded AI: Bot logic stays inside the Bot or Tank class.
2. Autopilot Class (Strategy Pattern): A specialized class that handles AI decisions and provides inputs (up, down, left, right, fire).
3. State Machine: A more formal state machine for bot behavior.

Decision Outcome

Chosen option: Autopilot Class (Strategy Pattern) (implemented in commit 1839288), because it decouples the "brain" (decisions) from the "body" (actions) and allows for more sophisticated AI logic without cluttering the entity classes.

Consequences

• Positive: Clean separation of concerns, easier bot testing, and improved AI behavior (e.g., pathfinding).
• Negative: Adds a layer of abstraction and requires careful interface design.
• Integration: Works with dependency injection (ADR-012) to access the game state.