Search control in an unknown environment using shortest path calculation and Lyapunov technique

Search control in an unknown space is a basic function of mobile autonomous robots. For search control, robustness against environmental change is important because environmental information is measured and updated at every moment. Robustness can be guaranteed by feedback controllers generated by global control Lyapunov functions (CLFs). To endow search control with the robustness properties of CLFs, Akiba et al. proposed search control based on a Lyapunov technique in an unknown space using a Q-learning algorithm. They showed that the robot reaches its destination with conventional control even if the destination is unknown. However, conventional control requires an offline calculation because the Q-learning algorithm has high calculation costs; hence, an online search is not achieved. Further, conventional control cannot be applied to complex environments such as those where humans live. In this research, we propose a Lyapunov based on search control in an unknown plane space. The proposed control employs Dijkstra's algorithm, and has lower calculation costs than conventional search control. Further, we successfully apply our method to mapping data in real environments via a simulation experiment.