A methodology for training neural-network agents to play a commercial-standard VALORANT-like 2v2 tactical shooter that runs on modest CPU at inference by replacing expensive pixel-based perception with a small set of ray-cast sensors that capture the essential spatial layout. The networks are trained by imitation learning (supervised on human trajectory data) so the agents reproduce human-like rather than performance-maximizing behavior, and human evaluations confirm the bots feel human while meeting the tight compute budget commercial games leave after 3D rendering.
Artificial intelligence (AI) has enabled agents to master complex video games, from first-person shooters like Counter-Strike to real-time strategy games such as StarCraft II and racing games like Gr…