Intelligent Control for Autonomous Mobile Robot based on Information Criterion on Virtual Environmental

Abstract

This paper proposes a new intelligent control method for autonomous mobile robots to acquire appropriate actions based on a real-time action search. In the proposed method, information criterion on the virtual environment is defined based on Kullback-Leibler divergence, which measures quality of the environmental information used for the action search. The robot searches suitable actions based on the environmental information whose quality is improved step by step. By applying the proposed method, the trade-off between calculation amount and information quality is taken into account, and the search process converges faster. The proposed method is applied to the moving obstacles avoidance problem, and its usefulness is verified through some simulation results.

Self‐position estimation of an autonomous mobile robot with variable processing time

“…Thus, in the current research we adjust the processing time to the quality of the environment information used in self-position estimation, in accordance with the concepts presented in Ref. 7.…”

“…For instance, in the physical resolution of a camera, or in the avoidance problem with multiple moving obstacles, which is considered as a specific example in Ref. 7, the moving obstacles in the environment are judged to be fewer than they actually are. When the camera resolution is low, the physical sensor processing time is short, but the accuracy of the resulting information about the environment itself is low, and as a result there are limits on the properties of the actions obtained.…”

“…In view of the fact that setting the situation recognition time in terms of ex- pected quality of the action is difficult, we proposed a method for representing the quality of the environment information used in the search so that the situation recognition time can be chosen appropriately to the information quality itself used in the search described in Ref. 7. Thus, in the current research we adjust the processing time to the quality of the environment information used in self-position estimation, in accordance with the concepts presented in Ref.…”

“…Furthermore, the time available for action acquisition varies depending on the situation of the robot's environment. Thus, we introduced the anytime computation [3] framework, and created a real-time action search method [6] that takes account of changes in the action acquisition time, and also a method of representing [7] the environment information and its quality that takes account of changes in the action search time. Action acquisition by the robot requires that it performs recognition of the situation currently encountered, and also action search using environment information acquired by situation recognition.…”

Hierarchical memory structure for the real-time search for action acquisition of an autonomous mobile robot

“…Fig.5 shows the motion planning system for obstacles' avoidance which incorporates the memory system. This system is constructed based on the system proposed in the previous work [5]. In this system, the robot's action is acquired through following processes.…”