| Summary: | 博士 === 國立交通大學 === 資訊科學與工程研究所 === 107 === Sensorimotor control is an essential mechanism for human motions, from involuntary reflex actions to intentional motor skill learning, such as walking, skiing, and swimming. By integrating perception from different sensory system, such as visual, vestibular, and proprioceptive system, and task goals, humans make decisions and generate corresponding motor commands to perform various motions. The sensorimotor control system of human has exceptional capabilities to perform skillful actions and represent the motor skills underneath different motions. Therefore, we can exploit the sensorimotor control model to simulate motions in a physically and biologically meaningful way.
The first goal of this thesis is to generate more natural responses closer to human reactions by integrating the effects of human perceptions on motor skills. We propose a sensorimotor control framework to simulate a preception-aware motion, motion sickness. Dynamic balance control is a fundamental motor skill of human beings based on the spatial orientation detected by the sensory system. Our framework, consisting of a balance controller and a vestibular model, can generate motion sickness like humans while receiving inaccurate estimations of spatial orientations. Our results demonstrate that sensorimotor control, integrating human perception and physics-based control, offers considerable potential for providing more human-like behaviors, especially for perceptual illusions of human beings, including visual, proprioceptive, and tactile sensations.
Next, we employ the sports techniques used by athletes to generate agile motor skills without reference motions. Skiing simulation involves many complex motor skills and physics, such as balance keeping, movement coordination, and articulated body dynamics. The techniques of inclination, angulation, and weighting/unweighting, obtained by kinetic analysis help design the simulation approach. These techniques can be the guidance of motion planning to compute physically-plausible control objectives to simulate carving turns and bump skiing with different speeds, turn radii, and bumps without reference motions. Moreover, we also employ the ski-snow contact model to compute snow reaction forces, which has been less considered in the computer animation, for counterbalancing the large variations of centripetal accelerations due to turning direction change.
In this thesis, we developed character animation approaches that utilize sensorimotor control to perform motor skills and respect the effects of perception to motions. The motions synthesized by our approaches are not only physically correct but also realistically natural since they are biologically inspired. Our approaches can replicate the mechanism of human motion generation and help understand the extraordinary abilities to execute various highly-skilled motions
|
|---|
