The purpose of this study is to suggest a proper tread of stairs using kinematic factors and moments of the lower-limb joints in the stair decent with the 3 different treads with boimechanical method in ergonomics. 9 subjects (body masses; 59.41±7.49, 64.03±6.65, 67.26:±7.58, heights; 160.50±6.35 ages; 31.22±2.99; parity; 1.67±0.71) participated in three experiments that were divided by physiological symptoms (the early (0-15 weeks), middle (16-27 weeks) and last (18-39 weeks). and they walked at self-selected pace on 4 staircases 3 trials. As extending the pregnancy period, cadence was shorter but cycle time were longer more and more. As extending treads of stair decent during pregnancy, speed, stride lengths and cycle time were increased. As extending the treads of stair decent, hip and ankle moments increased but knee moments decreased in sagittal plane. There were increasing or decreasing of moments by means of treads. These changes may account for relation between the treads of stair and moments in pregnant women. The main changes of pregnant women were joint moments and kinematic factors during pregnancy period because pregnancy makes them physical changes. It is possible that joints have connection with compensation each other to maximize stability and to control gait motion. In conclusion, we suggest that the tread of stair is longer than 26cm tread. and exercise programs to improve muscle activity were necessary where joint moments were small.

We present the synergy effect of humanoid robot walking down on a slope and support vector machines in this paper. The biped robot architecture is highly suitable for the working in the human environment due to its advantages in obstacle avoidance and ability to be employed as human substitutes. But the complex dynamics in the robot and ground makes robot control difficult. The trajectory of the zero moment point (ZMP) in a biped walking robot is an important criterion used for the balance of the walking robots. The ZMP trajectory as dynamic stability of motion will be handled by support vector machines (SVM). Three kinds of kernels are also employed, and each result from these kernels is compared to one another.