Design of machine game strategy for proactive collaboration with a human based on intent inferring and cognition identification

As a powerful analytical tool, human–machine games play an important role in numerous scenarios. However, the human decision-making process is often constrained by limited cognitive ability, which leads to their behavior deviating from the game equilibrium and affects the effectiveness of the machine's decision. Given this, designing appropriate game strategies for machines in human–machine games to enhance their cooperative intelligence is a challenging issue. This paper addresses this issue in the context of human–machine differential games. To this end, a linear matrix inequality optimization-based human intent inference method in an interactive setting is proposed, which enables the machine to compute human policies corresponding to different cognitive levels. Subsequently, using the computed human non-equilibrium policy model, a human cognitive level identification method along with a novel event-triggered detection mechanism is presented to detect the human cognitive level switching moments and the corresponding cognitive level during human–machine interaction. Then, based on switching control theory, a method for designing game control strategies for machines in human–machine games is proposed. Finally, a simulation experiment on a driver assistance system verifies the effectiveness of the proposed method.