Toward fully autonomous vehicle navigation: From behavioral to hybrid multi-controller architectures

This paper makes the focus on the way to increase gradually the autonomy of single vehicle as well as multi-vehicle systems (MVS) to achieve autonomous navigation in complex environments (e.g., cluttered, uncertain and/or dynamic). Its main objective is to give an overview of the developed generic control architectures (mainly decision/action aspects), and their different components, in order to enhance the safety, flexibility and the reliability of autonomous navigation. Furthermore, among the main ideas developed in this paper are related to the potentiality of using multi-controller architectures and their hybrid nature (continuous/discrete). Indeed, using this kind of control allows us to break the complexity of the overall tasks to be carried out while using bottom-up construction. This implies the development of appropriate reliable elementary controllers (obstacle avoidance, target reaching/tracking, formation maintaining, etc.), but also the proposition of specific mechanisms to manage the controllers' interactions, which ensures the respect of different constraints and enhancing metrics/criteria linked to the overall control reliability. Although the developed concepts/methods/architectures could be applied for different domains (such as service robotics or agriculture), the transportation-related purposes remains the privileged target. Two main examples of task achievement will be highlighted in this paper: appropriate flexible and reliable navigation based on sequential target reaching, and dynamic multi-vehicle navigation in formation.