Qianqian Zhang

Water is a basic human necessity, and the amount of water on Earth remains fairly constant

In recent years, frequent floods hit Chinese cities and caused heavy casualties and property losses, making China faced with severe flood problems. In this study, Nanhai Future City in the IX Flood Control Area of Yancheng City, Jiangsu Province, was selected as the research area to simulate water-level changes under different control schemes meeting extreme rainfalls. MIKE model simulated the inundation with the designed storm of different return periods. The results showed that flooding inside the research area was severe. Higher drainage capacity of the pump stations with more engineering and non-engineering measures can reduce the adverse effect of extreme rainfall. The results provide a reference for planning future infrastructure and flood control decisions for Nanhai Future City and the surrounding areas.

In order to analyze the influence of the objective function on the algorithm, we compare the common definitions of objective functions and propose a piecewise objective function definition method based on error value to further improve the algorithm performance. Using classical swarm intelligence optimization algorithms such as DE, MABC, PSO, three types of objective functions are contrast analyzed on a 6-degree-of-freedom (DOF) underwater manipulator model to verify the superiority of the improved objective function. The results show that the improved objective function performs well in solving the inverse kinematic of the 6-DOF underwater manipulator, which is reflected in that the objective function is insensitive to the algorithm and the rod of the manipulator. That is, the improved objective function is more stable, and the proportion of the posture error to the specified accuracy is higher. The objective function definition method of adaptive weight proposed in this paper is an effective and suitable definition method for multi-DOF manipulators.

Wearable activity recognition is an important area for healthcare applications, especially for children and elderly people. Current research proves that deep learning techniques are capable of learning deep features from the sensor raw signal. However, this task faces two challenges. First, it is hard to capture the dynamic and non-linear interactions between different sensor modalities and time slots. Besides, the spatial relationship between sensors convey significant information for wearable activity recognition, thus how to model the sensor relationship is one of the problems that need to be solved. In this work, we propose GraphConvLSTM, a scalable framework that model spatial relationship, local interaction, and long-term temporal dependency simultaneously. We evaluate our model on two real-world activity recognition datasets, and it achieves 1.75%, 2.68% accuracy improvements over existing methods (accuracy: 90.83%, 93.08%) for wearable activity recognition task, which demonstrates the ability of the proposed approach to learn spatial, local and global temporal features.