Real Time Data Logging And Status Analysis For Autonomous Ground Rodot System

The autonomous ground robot (AGR) is a sophisticated system capable of intelligent decision making and task performing. Generally, its navigation pipeline comprise of several connected parts such as sensor signal acquiring, data processing, information fusion, behavior shaping, and motion controlling. Because of the distributed nature of pipeline, typically, there are multiple intelligent agents concurrently running. Therefore, it is of a great challenge to explore the problem of logging and analyzing the real-time and distributed data stream in the navigation procedure.To improve the understanding of above-mentioned problem, this thesis firstly provides a comprehensive survey on the previous works of AGR architecture, data logging, and status monitoring. In particular, two typical messaging architecture, ROBIX and NML, are elaborately depicted as all experiments of this research are based on them. Secondly, a concise design of key modules of logging system is introduced for AGR, including communicating, storing, sharing, threading, and journaling. Experimental results of performing testing are also presented. Lastly, in the status analysis part, the adjust autonomy problems are explored. Specifically, sequential data from sensors are used to determine the autonomy mode under a Multi-armed Bandit framework which contains an online Bayesian algorithm to estimate the optimal autonomy score continually. The experimental results of the autonomy mode decision are encouraging.