Trust as a metric for auction-based task assignment in a cooperative team of robots with heterogeneous capabilities

As robots become part of our everyday lives, they may be required to cooperate without being aware of each other's capabilities (e.g., because different teams have developed them), and therefore will have to trust each other to work together safely and efficiently. Starting from this premise, this work identifies trust as an essential metric to assign tasks to robots using auction-based mechanisms. We model trust by taking inspiration from popular models in the literature and adapting them to an open environment in which heterogeneous robots may dynamically enter or exit, execute assigned tasks, or verify the correct execution of tasks by other robots. Robots are considered to be heterogeneous in the sense that they may have different capabilities in executing and verifying the execution of actions. In the proposed model, “doing an action” and “verifying the execution of an action” are distinct, not necessarily overlapping, capabilities. Some robots may be able to do an action, whereas some robots may not be able to do it but only to observe and judge the ability of other robots to do it. After introducing the relevant formalism, the article describes the system's architecture implemented in ROS and multiple experiments performed in simulation and with real robots (one NAO and two Pepper robots by SoftBank Robotics), providing a proof-of-concept for broader utilization of the system in cooperative robotic scenarios.