ABSTRACT

Continuous Hybrid HMM .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85 4.6 Development and Manipulation of Proto-Symbols Based on Geometric

Proto-Symbol Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88 4.6.1 Hierarchical Mimesis Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88 4.6.2 Definition of Distance Between HMMs . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 4.6.3 Construction of Space Based on Similarity . . . . . . . . . . . . . . . . . . . . . . . 90

4.7 Behavior Manipulation by Proto-Symbol Manipulation . . . . . . . . . . . . . . . . . . . . 90 4.7.1 Proto-Symbol Manipulation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90 4.7.2 Generation of Novel Proto-Symbol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

4.7.2.1 Novel Motion Generation from State Sequence in Proto-Symbol Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

4.7.3 Recognition of Novel Motion Based on Proto-Symbol Manipulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 4.7.3.1 Novel Motion Recognition as State Sequence in

Proto-Symbol Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96

4.8 Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 4.8.1 Recognition of Novel Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 4.8.2 Generation of Novel Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

4.9 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99 4.9.1 Comparison with Conventional Research . . . . . . . . . . . . . . . . . . . . . . . . . 99

4.10 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 Acknowledgment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

The bodies of humanoid robots recently show strides of advancement by adopting cutting-edge mechatronics and manufacturing technologies. This fact forms the background of the conducted research. It is necessary to found the technology to design a comparable information processing system. Namely, for building the brain of a humanoid, it has turned out that there are many essential problems that seem unsolvable based on the conventional analytic approaches of robotics. This research aimed to approach the essential problems of machine intelligence such as the whole-body motion pattern generation of large DOF systems, the dynamic interaction between the body and the world, embodied symbol emergence, and intelligence through emerged symbol manipulation, by carefully investigating and adopting the paradigms of neuroscience in a constructive way.