arXiv cs.AITuesday · May 26, 2026FREE

Quantum Frog: Emergent Cooperation and Difficulty Scaling in a Quantized-Time Cooperative Game

A new paper from arXiv cs.AI, published on May 26, 2026, details "Quantum Frog," a two-player cooperative game featuring a novel quantized-time mechanic. In this system, the game environment progresses only when a player executes an action, a departure from real-time or turn-based systems. The game, an 8x8 grid inspired by Frogger, requires two frog agents to cooperatively navigate traffic and reach the opposite side simultaneously. Researchers employed reinforcement learning (RL) to analyze four key design questions: how game difficulty scales with traffic density, the optimal single-agent policy, the cooperation gap between independent and cooperative multi-agent play, and the emergent joint strategies. Agents were trained through five escalating stages, utilizing methods such as Tabular Q-Learning, Deep Q-Network (DQN), Independent DQN (IDQN), and Multi-Agent Proximal Policy Optimisation (MAPPO with a centralized critic). These models were evaluated across traffic densities ranging from one to six cars. A primary finding was that the quantized-time mechanic inherently promotes a "rush strategy," where agents consistently move directly upward. This strategy proved universally optimal because it minimizes the agents' exposure time to environmental hazards, simplifying the decision-making process and highlighting a unique interaction between game mechanics and emergent AI behavior.