This paper presents novel RNN structures. In order to reduce the calculation as well as to overcome the analysis difficulty of gate structure and the problem of data dependence related normalization in the conventional method, 4 types of new RNNs inspired by shortcut connection were constructed while taking the advantages of gated RNNs and a sech function. Experiments and analyses are carried out to evaluate the performance of the RNNs. In detail, the experiments are performed using two corpus of WikiText-2 and IMDB, which have different properties. As a result, in the binary classification task using IMDB, the accuracy was about the same as LSTM and GRU with the parameter of 1/6 or less. However, in the language modeling task using WikiText-2, even if multilayer and the parameters of the intermediate layer were increased, the results were worse than that of conventional RNNs. This should be clarified in our future research.

This paper describes comparative analyses of time-efficient multi-arm assembly using a task board for a belt drive unit. The task board was used for the World Robot Summit (WRS) 2018 competition. From the viewpoint of using multiple arms, handling multiple parts simultaneously, and representing assembly both by robot systems and by human subjects, this paper introduces the set that shows only the contact condition between the arm and the part to arrange the time series data easily. Based on this analysis, there is a significant difference in the whole work time between the robot systems and the human subjects. One of the reasons is the difference in the usage of multiple arms. It will be important for robot systems to utilize multiple arms simultaneously to realize the time-efficient multi-arm assembly.

For users to carry out various jobs according to the construction plan, if unexpected machine failures occur and their machines go down for an extended time, they will be huge losses. Therefore, machine maintenances are required for their machines to prevent from machine failures. However, due to operation in severe environment condition such as high load or long time use and in unexpected use, they often fail earlier than expected. For the maintenance of construction machinery, we propose to detect early indications of failure by predicting remaining useful times. Thereby, their machines can be performed maintenance before their failures and prevent unexpected machine failures. We propose to predict the machine failures of lower traveling bodies of hydraulic excavators by estimating their remaining useful times. Moreover, we also propose a practical example of maintenance activity using remaining useful times prediction in addition to failure prediction by neural network for hydraulic excavators and its effectiveness is shown.

In a modern aging society, electric wheelchairs are an essential means of transportation for the elderly. In contrast, human operation errors of the electric wheelchair cause some accidents. For a control system operated by a human, human assist control to avoid accidents attracts much attention in recent years. For this problem, Hayashi et al. have proposed a human assist control of the electric wheelchair by using the control barrier function. However, the proposed method had a problem that delays human assist control due to a Laser Range Finder which needs some time to get the distance to obstacles. In this paper, we consider a depth sensor as a novel position measurement sensor. The depth sensor attracts much attention because the depth sensor is cheap and high availability. We evaluate the effectiveness of the depth sensor for human assist control of the electric wheelchair. Finally, we demonstrate the superior performance of the depth sensor by actual experiments, compared with a Laser Range Finder.

Deformation in skeletal muscle during contraction is restricted by skin, bone, and adjacent muscles. In this study, the relationship between these constraints and muscle function was examined using finite element simulation. An unconstrained model (UC), a model with limited bulge in the direction of skin and bone (CC1), and a model with limited bulge in the direction of skin and bone and adjacent muscles (CC2) were used. In comparison to UC, the muscle strength in force-length relationship decreased by 6.2% in CC1 and 9.9% in CC2 on average. It suggested that the difference in strain distribution of muscle tissue and the energy loss other than the direction of the force line due to bulging and rotation of muscle resulted in these force differences.