TacticCraft: Natural Language-Driven Tactical Adaptation for StarCraft II
This work addresses the need for flexible tactical control in complex real-time strategy games like StarCraft II, offering practical strategy customization for AI agents, though it is incremental as it builds on existing pre-trained policies.
The paper tackles the problem of StarCraft II AI agents lacking adaptability to high-level tactical directives by introducing an adapter-based approach that conditions on strategic preferences, resulting in successful modulation of agent behavior across tactical dimensions while maintaining competitive performance with minimal computational overhead.
We present an adapter-based approach for tactical conditioning of StarCraft II AI agents. Current agents, while powerful, lack the ability to adapt their strategies based on high-level tactical directives. Our method freezes a pre-trained policy network (DI-Star) and attaches lightweight adapter modules to each action head, conditioned on a tactical tensor that encodes strategic preferences. By training these adapters with KL divergence constraints, we ensure the policy maintains core competencies while exhibiting tactical variations. Experimental results show our approach successfully modulates agent behavior across tactical dimensions including aggression, expansion patterns, and technology preferences, while maintaining competitive performance. Our method enables flexible tactical control with minimal computational overhead, offering practical strategy customization for complex real-time strategy games.