Robot Training Under Conditions of Incomplete Information
This work addresses robotic system training for scenarios with data limitations, but it appears incremental as it builds on prior adaptive algorithms without introducing a fundamentally new approach.
The paper tackles the problem of training robotic systems under incomplete information by using an operator's decisions to adaptively form the robot's behavior, resulting in flexible responses to environmental instability through switching between preset models.
The development of the works of the author about adaptive algorithms of teaching the robotic systems with the help of operator is described here. An operator is assumed to be an experience decision-maker and sane carrier of a target which the robotic system needs to achieve. The works characteristic is that the behavior of the robotic system is not specified a priori (as standard) but is formed adaptively based on the information about the situation and decisions made by a decision-maker. In this scheme the robotic system and the decision-maker can cooperate in the normal operation mode of the robotic system or in the time sharing mode with the possibility to plan actively the experiment on the robotic system. If the adaptive scheme is chosen, there are teaching stages and operating stages of the robotic system. At that the decision-maker can act slowly having the possibility to weigh the decision made. This way allows the robotic system reacting flexibly by switching between preset models and respond to the environment instability. The data integrity about the environment condition and about target preferences of an operator plays a very important role in robotic system work. The effective work of the robotic system depends on the effective settings of a preference model of the robotic system based on the decisions of the decision-maker and on the effective control. The influence of settings and control factors on the index of effectiveness of the robotic system is subject of this work. The uncertainty may be caused by the data flow limitation received by the operator on the stage of the model setting.