GSPN (Generalized Stochastic Petri Nets) based navigation behavior control of an indoor service robot

In this paper, we propose a behavior selection scheme between a reactive motion control combined with a global path planning algorithm and combined with a trajectory tracking algorithm. We design two navigation behaviors. One is a plan-based reactive navigation behavior, the AutoMove_reactive, which exploits the conventional Dynamic Window Approach (DWA) as the reactive motion controller. The other behavior is AutoMove_tracking, which exploits trajectory tracking controller. Two navigation schemes are based on different navigation assumptions. As a result, the two schemes have completely different advantages and disadvantages. We selected the appropriate navigation behavior based on the Generalized Stochastic Petri Nets (GSPN) discrete control framework. The main problem in switching between the behaviors is as follows: the determination of when the behavior should be switched and the method for determining the switching behavior conditions. One of the major drawbacks of DWA is the local minimum problem. If a robot falls into the local minimum, the navigation performance decreases degrades greatly. Therefore, a robot should switch its navigation scheme when it detects the local minimum. Experimental results showed that the proposed model was more effective than that based on the using single navigation technique. The AutoMove_reactive is mostly selected when obstacles are moving around the robot. The AutoMove _reactive local minimum situation due to dynamic obstacles or static obstacles was suitably managed using the GSPN. The AutoMove_tracking showed reliable tracking when the robot entered a narrow corridor. Moreover, AutoMove_tracking can recover from the AutoMove_reactive local trap situation. The resultant navigation performance improved from the performance achieved with the single navigation behavior.