抄録
In this paper, we have proposed a behavior selection framework using GSPNs (Generalized Stochastic Petri-nets). We exploited two navigation schemes, a sensor-based reactive scheme, viz., DWA, and map-based tracking using trajectory tracking. Our selection framework is distinct from previous behavior selection schemes because the exploited behaviors are able to ensure the completion of navigation tasks. We designed two sub-systems to monitor the states of each navigation scheme. The proposed selection framework in this study was simulated using the Player/Stage simulator. We carried out performance estimation of the two exploited navigation schemes. The results of performance estimation clearly showed that both navigation schemes are required for robust and efficient navigation in real-world environments. The average navigation time under the proposed behavior selection framework decreased by about 45% in comparison with the use of a single navigation scheme. These results clearly showed that our behavior selection framework is more efficient and robust than a single navigation scheme.
