Hybrid Control for Robot Navigation - A Hierarchical Q-Learning Algorithm

Autonomous mobile robots have been widely studied and applied not only as a test bed to academically demonstrate the achievement of artificial intelligence but also as an essential component of industrial and home automation. Mobile robots have many potential applications in routine or dangerous tasks such as delivery of supplies in hospitals, cleaning of offices, and operations in a nuclear plant. One of the fundamental and critical research areas in mobile robotics is navigation, which generally includes local navigation and global navigation. Local navigation, often called reactive control, learns or plans the local paths using the current sensory inputs without prior complete knowledge of the environment. Global navigation, often called deliberate control, learns or plans the global paths based on a relatively abstract and complete knowledge about the environment. In this article, hybrid control architecture is conceived via combining reactive and deliberate control using a hierarchical Q-learning (HQL) algorithm.