Autonomous mobile systems (AMS) rely on multiple objects and intelligent inference that are able to take appropriate actions even in unforeseen circumstances. It is challenge even for the development of a single AMS due to the complexity of the system, uncertainty of the environment, and multiple aspects of the cooperation within the system. It is more challenge for the bipedal robots due to the balance and stable movement brought from the structure. Thus it is more challenge for the multiple robotics systems (MRSs). Due to the numerous amount of application areas of various types of robotics system, the requirements and research of robot application are increased rapidly. Multiple robots are widely adopted in the space exploration, field searching, agriculture as well as many other application. Coordination and collaboration become one of the main concerns for the MRSs.
To build a trustworthy architecture of MRS with the various complexity and human factors, it is important to consider the stability and mobility of a hierarchical structure. In this work, we introduce the agent concept to handle task coordination. The proposed general framework of a trustworthy MRS task coordination allows the development of complex multi-robot behavior via hierarchical and sequential composition of control interface and estimation arguments, and parallel composition of agents.
At the highest level of the hierarchy, the multi-robot system is represented by three interacting agents: a coordination agent, a task agent, and a task net. The coordination agent reduces to the specification of communication channels between task agents, and the specification of parameters for transitions and the instantiation of each agent within the task agent among robots.
Each task agent carries the behavior description specified by task net. A task net is an individual or composite net that is represented by a PrT nets, which will be described in next section. A task net can describe the sequential and/or concurrent tasks by the defined composition rule, which are regulated based on the transition and guards. This is schematically illustrated in Figure 1.
Task agents can receive estimates of the obstacles from other robots, and commands and specifications from the human operator on input channels, and it can send its own information to other robots or to the human operator on the output channels.
For verification purpose, in general we introduce the notion of temporal constraints for transitions, places and arcs. That is, a place may hold one token for a certain marking. Otherwise, the formal analysis would become more cumbersome. This is a trade-off between expressiveness and analysis power.
Faculty: Dr. Yujian Fu, AAMU