This paper presents a method to learn stepping motions for fall avoidance by reinforcement learning. In order to overcome the curse of dimensionality associated with the large number of degrees of freedom with a humanoid robot, we consider learning on a reduced dimension state space based on a simplified inverted pendulum model. The proposed method is applied to a humanoid robot in numerical simulations, and simulation results demonstrate the feasibility of the proposed method as a mean to acquire appropriate stepping motions in order to avoid falling due to external perturbations.
In this paper, a pneumatically driven forceps manipulator which is suitable for suturing task is developed. The manipulator has a wire drive rotating joint at the tip part and enables precise gripper rotation into practice. The gripper mechanism which consists of a small cylinder and slider-crank mechanism is proposed and implemented at the tip part. The gripper is driven by air pressure supplied through a fine tube so that the interference of other joints can be avoided. The gripping force of 20[N] is achieved. We constructed a master-slave system with the forceps manipulator and conducted some suturing experiments using a sponge object. It became clear that the forceps manipulator had good maneuverability and could estimate the external moment within the uncertainty of 10[mNm].
This paper is concerned with trajectory tracking and obstacle avoidance control using avoidance manipulability of redundant manipulators. We have proposed a new manufacturing system using robot to deal with object with unknown shape by combining an avoidance control system and preview control system. However, through analyses and simulation studies it develops that the shape of redundant manipulator was not always kept to the best configuration on the view point of shape-changing ability. In this paper we propose a new criterion to evaluate the shape-changing ability in the configuration space while tracking the hand-desired trajectory. Using this criterion we constructed real-time configuration control system with preview evaluation by introducing imaginary manipulator in future time. Finally the proposed system was evaluated by several simulations on the point of real-time configuration optimization, and the feasibility of total system was analyzed.
It is important to make its vision system more robust and accurate, to give optimal visual-feedback, which helps to control a robot. We propose a robust and accurate pattern matching method for simultaneously identifying robots and estimating their orientations that does not use color segmentation. To search for similar patterns, our approach uses continuous DP matching, which is obtained by scanning at an ellipse circumference from the center of the robot. The DP similarity value is used to identify object, and to obtain the optimal route by back tracing to estimate its orientation. We found that our system’s ability to identify objects was robust to variation in light conditions. This is because it can take advantage of the changes in intensity only.
Humans’ primitive skill of imitative learning is regarded as an origin of human intelligence because it is said that imitation is fundamental function for communication and symbol manipulation. M.Donald has proposed “mimesis model” in order to approach to a symbol emergence framework from behavior recognition/generation for humanoid robots. In this paper, we propose a mathematical model for motion recognition and generation as combination of basic motions by proto-symbol manipulation which is abstract expression of motion patterns. In order to describe the proto-symbol manipulation as geometric manipulation, construction of proto-symbol space and geometric proto-symbol manipulation method are established.
We have applied a parametric excitation method to a kneed biped robot with semicircular feet and have shown that the robot can walk sustainably with only knee torque. A swing-leg of the kneed biped robot has similar mechanism to an acrobot, and many acrobots are controlled in inverse direction like ornithoid walking. These suggest that inverse bending of a knee restores more mechanical energy than forward bending, and hence, the ornithoid walking can be more efficient. In this paper, we first compare the forward bending with the inverse bending for a double pendulum, and show by numerical simulation that the mechanical energy of the inverse bending increases more than that of the forward bending like human walking. We then propose a parametric excitation based ornithoid gait for a kneed biped robot, and show sustainably walking by numerical simulation. Finally, we compare parametric excitation based ornithoid gait with parametric excitation based human gait, and we show that ornithoid gait is more efficeint.
In the autonomous mobile robot using ultrasonic range finders, especially in outdoor environment, the ultrasonic range finder is required to detect the object frequently at the short distance for driving control and obstacle avoidance, and to find the landmark object at a long distance for navigation system. However the frequency and the long rang measurement at same time are incompatible in the conventional ultrasonic range finder. The aim of this paper is describe about the ultrasonic range finder that processes these two opposite functions in parallel. This ultrasonic range finder transmits wave coded the pattern that is composed in transmitting patterns of each functions, then calculates cross-correlation between transmitted pattern and received signal at each functions. The experimental results show the ultrasonic range finder measures distance to the short range object frequently and the long range object at same time.