The Internet of Things is an emerging worldwide ecosystem in which smart devices interact to build smart homes, smart cities, etc. In such context, plethora of research efforts are oriented to networking and devices; analyzing and formalizing IoT systems are still in their earliest state. In this paper, we propose a BiAgents* (Bigraphical Agents) model to provide a formal description for IoT systems structure and behavior. In addition, we encode the BiAgents* specification into Maude language to enable an autonomic execution of the IoT systems' behaviors. The proposed approach is illustrated and evaluated through an example (an intelligent case of collision avoidance system).