Many researchers have developed musculoskeletal robots with McKibben pneumatic actuator to elucidate the movement mechanism of animals by a constructive approach in previous studies. These robots can realize various dynamic motions by simple control, such as adjustment of the pressure input patterns. However, the pressure input patterns are often determined by trial and error, and it has not been clarified what kind of mechanism an autonomous coordination pattern like animals is generated. In this study, we propose a new control law based on some force feedback to achieve autonomous coordination of antagonistic muscles. The control law is applied to a leg model with asymmetric antagonistic structure, and the capability was analyzed through numerical simulations and experiments with an actual robot. From these results, it is confirmed that the proposed control law can generate cooperations of the antagonistic muscles and realize periodic motion of the leg.

In this paper, we aim to evaluate an occupancy lighting control by using vision-based motion sensor compared with passive infrared (PIR) motion sensor. Our results showed that vision-based sensor improved detection accuracy by 21.7% and detected small movement of occupant more accurately than PIR sensor. We also confirmed that vision-based sensor enabled to reduce off-delay, which is introduced to prevent undesired switching off light, by 7.92 min (74.8%). As a result of simulation for occupancy lighting control with vision-based sensor and PIR sensor, vision-based sensor enabled to improve energy saving rate by 8.9%. Additionally, we discussed disadvantages of vision-based sensor confirmed in our experiment and considered solutions for them.

In this paper, we propose a method to detect workers’ nonstandard motion in cell production lines. Our method consists of three features: (1) sensing motion of a person, (2) measuring the time of motion by using Dynamic Time Warping and (3) analysis of the motion time. We applied the method to a prototype cell production line in our factory and verified its effectiveness. In an interview with Industrial Engineer, we have confirmed that our system has the potential to reduce the time taken for discovering nonstandard motion from 10 hours to 2.5 hours. From the above, we conclude that our method can enhance the efficiency of cell production line improvement.

Neural networks have high performance in tasks such as image recognition. However, the computational cost is high, and it is difficult to implement them on small devices. In recent years, in order to solve this problem, research on compression techniques of the neural network has been advanced. “Pruning” is known as one of the important approaches for compression techniques. A “sensitivity map” is a map that visualizes which area of the input data the model focused on. However, it has not been analyzed much on how the model structual changes by pruning affects the sensitivity maps. In this paper, we analyze the influence of pruning on the sensitivity maps. As a result, it was found that the region of interest in the background of the sensitivity map was reduced after pruning.

This study proposes a method to apply deep neural networks to controllers of robotic swarms. In a typical approach to design controllers, the designer has to define the features extracted from sensory inputs in advance. By applying deep neural networks with convolution layers, it can automatically extract features from sensory inputs. We applied two methods to train the deep neural networks, i.e.,deep reinforcement learning and deep neuroevolution. The controllers were tested in a path-formation task using computer simulations. Compared with deep reinforcement learning, deep neuroevolution was able to generate collective behavior even in sparse reward settings.