Probability-Based Context-Aware Robotic System

The development of ambient intelligence in the past decades calls for the smart environment that is able to respond to entities inside. The technology of context-awareness is a promising solution to this mobile distributed computing, not only in small personal electronic devices, but also in robotic systems.A prototype of context-aware assistive robotic system is designed and implemented in this thesis on the purpose of providing assistive services to the seniors and people with disabilities. The issue of context reasoning is extensively reviewed and researched. The particle filtering algorithm is improved in terms of proactive particle sampling, fuzzy based velocity estimation, and active observation selection, for the basic tasks of robotic tracking and following. The Monte Carlo partially observable Markov decision process is enhanced in terms of belief state augmentation, value function computation based on direct sampling, and non-greedy action execution, for the task of path planning. Experimental results are carried out to validate the feasibility and effectiveness of the proposed solutions. The prototyped robotic system can be further implemented on other potential applications for the aim of advanced service provision.