Exploring different cooperation strategies in nonzero-sum hybrid games

Cyber-phyiscal systems (CPS) like trains, planes, autonomous cars or robots need verification to ensure their safety. Especially relevant are situations involving two CPSs, as these situations occur frequently in real-world. The challenging aspects here are the possible interactions between the CPSs that arise from their intentions or goals. The logic dGLsc which has been developed in our group, deals with these situations by regarding them as games and the involved CPSs as players, each with an individual goal they try to reach. In order to achieve their goal, the two players may collaborate with each other when helpful and may compete when necessary. Therefore, dGLsc is able to naturally model interactions involving two CPSs by taking both of their goals into account, enabling an accurate verification of the situation.

In dGLsc, the players cooperate whenever possible, i.e. if a player has several options with which they can reach their goal, they will choose one that helps the other player reach theirs too. But how would the semantics change if players would for example, only cooperate when strictly necessary? How can players act with foresight if the game lasts more than one round? In this thesis, the semantics of dGLsc should be modified to incorporate different strategies for entering a cooperation, and possible combinations with dGLsc should be explored.

For more information, please reach out to julia.butte@kit.edu.