This paper proposes a method for temporal action spotting: the temporal segmentation and classification of human actions in videos. Naturally performed human actions often involve actor's unintentional motions. These unintentional motions yield false visual evidences in the videos, which are not related to the performed actions and degrade the performance of temporal action spotting. To deal with this problem, our proposed method empolys a voting-based approach in which the temporal relation between each action and its visual evidence is probabilistically modeled as a voting score function. Due to the approach, our method can robustly spot the target actions even when the actions involve several unintentional motions, because the effect of the false visual evidences yielded by the unintentional motions can be canceled by other visual evidences observed with the target actions. Experimental results showed that the proposed method is highly robust to the unintentional motions.

This paper investigates the hovering efficiency of the wing-actuation parameters of a flapping microaerial vehicle with a two-degree-of-freedom (2-DOF) driving mechanism, using finite-element analysis based on the arbitrary Lagrangian-Eulerian method (ALE-FEA). A 75-mm-long wing and multilinkage mechanism that consisted of thin plates and films was employed. It generated a flapping motion that consisted of horizontal stroke of the wing and twist around its leading edge, which were activated by the rotations of two motors. The application of the ALE-FEA was successfully extended to the structural behavior of the driving mechanism. The behaviors of the linkage mechanism, the wing, and its surrounding airflow were reproduced numerically. Various parameters were obtained, some of which were difficult to measure in real experiments, such as the pressure distribution on the wing. For cases of 25 combinations of wings' twist angles and flapping periods, lift forces and energy requirements for the motors were determined. Based on these results, the most efficient flapping combination that generated 1 gf lift force was calculated. In contrast with the combination that created the strongest lift force, the combination of a larger twist angle and a faster flapping cycle was found to provide the most efficient flight.

An actuator for a middle-sized flapping flight robot is quantitatively investigated. Flapping flight like that of insects is a potentially useful method of travel for micro robots. Some insect-mimicking robots have been developed, which have actuators with two or more degrees of freedom per wing. For the detailed design of such an actuator, it is necessary to deal with the interaction between the behavior of the actuator and the aerodynamic forces generated by the actuation. In particular, the torque and power requirements of each degree of freedom are essential in the choice of a suitable motor, but it is impossible to determine these without considering the interaction between the driving force of the actuator and the reaction force of the wing from the surrounding airflow. Here, we achieved an analysis of the mechanical aspects of flapping flight, actuator and wing, using finite element analysis based on the arbitrary Lagrangian-Eulerian method, which can treat the fluid-structure interaction problem properly. We worked out the spec of motors for the 2DOF actuator, and investigated the structural tolerance. We developed a useful tool for the design of flapping flight robots

We investigated some configurations of middle-size and high-aspect ratio wings for flapping flight by fluid-structure interaction simulation. Mimicking the flapping flight of insects is a useful method of microrobot motion. Its comprehensive unsteady aerodynamics has gradually been unraveled and several types of flapping-flight robots have been proposed. We examined the motion of a dragonfly, and try to apply it to microrobots. In designing such nimble flapping flight, adequate evaluation of the deformation of the wing is unavoidable. Wings of such species have high aspect ratio because a small momentum of inertia around the longitudinal axis is a great advantage in controlling. Such a slender wing is easy to twist around the longitudinal direction. Thus, in order to design such a wing, the wing behavior caused by interaction with the airflow must be analyzed, and the adequate stiffness must be determined. We designed two types of wings based on the architecture of the dragonfly’s wing, and examined the performance of the wings by fluid-structure interaction analysis. Here we show some examples of the designs and the performance of the wings for hovering, as the results of the first trial.

An insect's flapping flight is an attractive method of transportation. It shows many useful flight modes such as maneuverable turning and stable hovering. In recent studies, experimental methods and numerical simulations have been successfully applied to solve the mechanism quantitatively. However, it is extremely difficult to accurately estimate the deformation of the wing caused by interaction with airflow. In prior studies, we analyzed this problem using a novel numerical simulation, fluid-structure interaction analysis, in order to represent this interactive behavior accurately, and achieved the quantitative evaluation of dragonfly hovering with flexible wings. As an advanced analysis, we added a solution of the interaction between the body and wings to the method, and realized free-flight simulation. Here, we demonstrate a sharp turn of 10 rad/s in a numerical simulation using a model based on an actual dragonfly.