Mobile robot control architecture for reflexive avoidance of moving obstacles

In this paper, a three-layer (deliberative, sequencing, reflexive) architecture is adopted and the structure of the reflexive layer is discussed. The objective of this architecture is to extract the basic actions that require hard-real-time execution from non-real-time-allowed behaviors by separating them into the reflexive and sequencing layers, respectively. The reflexive layer consists of resources, actions, an action coordinator and a motion controller. To guarantee the hard-real-time execution, a set of simple actions and an action coordinator are designed using the functions provided in the RTAI (Real-Time Application Interface for Linux) environment. Also, an obstacle avoidance algorithm based upon data from a laser range scanner is developed. For the purpose of avoiding a moving obstacle, which is treated as a moving circle through segmentation and circularization processes, a Kalman filter is developed to estimate the distance and the heading of the center of the moving circle. The effectiveness and real-time characteristics of the proposed reflexive layer and the developed algorithms are examined through experiments using scattered stand-still obstacles as well as a moving human.