Information Fusion for Hypothesis Generation under Uncertain and Partial Information Access Situation. Updated 2008

摘要

The challenge for most artificial systems that operate autonomously in the real world environment is how to cope with dynamic environment with limited, uncertain, and noisy information. Artificial intelligence and intelligent robotics research has been trying to solve such a problem by either improving accuracy of recognition systems or by integrating multiple source of information. In addition, architectural issues has been discussed on whether classical Sense-Model-Plan-Act architecture or the subsumption architecture better suits for autonomous agents. Information fusion issue is tightly coupled with behavioural control as overall performance of the autonomous system is the ultimate concern. The work performed focused on identifying possible system architecture for realistic information fusion and corresponding reactions under uncertain environments. Our research starts from analysing issues in existing paradigm of autonomous agent and AI architectures, redefine needs, and propose a suitable architecture. In this research, it was essential to learn from biological systems where various species has evolved to adapt to uncertain and dynamic environment for survival.