Combining reactive and deterministic behaviours for mobile agricultural robots

Autonomous vehicles for agricultural applications should be able to perform complex tasks, which require planning and optimization before execution, as well as context-dependent reactive behaviour during execution in the field. Hybrid control architectures are well suited to such applications. In this paper, a formal task-modelling framework is proposed and implemented, based on such a hybrid architecture. Each task is modelled as a sequence of transitions among operational modes, and each mode is defined as an instance of a set of elementary object-behaviours, which execute concurrently. A case-study ‘navigation’ task was implemented, where deterministic path following and reactive obstacle avoidance behaviours were combined. The robot's operation was successfully tested in an open field and ‘smooth’ mode switching without chattering was observed in the presence of both static and dynamic obstacles. The virtual force field method based on a dynamic occupancy grid seems to be a very good candidate – in terms of smoothness of mode transitions – for implementing obstacle avoidance behaviour.