Zhou Fengyu

The design and application of human attitude detector based on inertial measurement unit

In order to decrease the casualties, monitor the real-time motion and safety of the fireman in fire scene, an accurate, efficient and low-cost attitude detector is designed. To monitor the attitude angle in real time, the extend Kalman filter for attitude detector based on quaternion is presented internally. To improve the accuracy detector uses the three-axis MEMS gyroscope with range of 1200 ° / s, and sensor data is processed by using a dynamic zero-calibration method and a configurable Parks-McClellan window FIR low-pass filter. The three-axis MEMS accelerometer and compass are also used. In consideration of the compensation of the incline the affect of soft and hard magnet, and the centripetal force, automatic calibration system is designed to be adapted in different environment to guarantee the dynamic correction. At the same time, detector integrates the wireless radio frequency technology and the OpenGL representation of the real-time human motion in 3D are also given. The results of experiment show that the heading and attitude data computed from the accelerometers and magnetometers is used for compensating the drift of gyroscopes, and the dynamic accuracy is controlled within 2 degree, meanwhile, the static accuracy is no more than 0.5 degree. Through the test of principle prototype, the feasibility and stability of the attitude detector are also demonstrated.

The establishment of movement model and the implementation of simulation control system for ship autopilot

A method is proposed in the paper that the parameters of a ship autopilot are inspected and regulated simulating the actual ship environment. Various math models are established during the process of ship movements. The main control software and SCM system used for the actual ship simulation are developed. The sea environment is simulated with the method of combining digital and physical simulation. A design project of ship simulating control system for autopilots is introduced here. The system shows great economic benefit to reduce the cost for manufacturing and maintaining autopilots.

Design of a Cloud Robotics Visual Platform

In recent years, the vision of robot has got great development. Because of the low sampling rate and the heavy calculation burden, the vision is obstructed to apply for robot. In order to solve the above problem, in this paper, we present the Cloud Robotics Visual Platform(CRVP) for the development of robot vision. The main contributions of the paper are as follows. (1) The parallel computational model of Map/Reduce is used to decrease the image time cycle of image training and recognition. (2) The image recognition engine is built with the Service-Oriented Architecture. (3) The H264 encoding algorithm and the Real time protocol are adopted to realize the video transmitting. Finally, the experiment is done using ECS of Aliyun. The results show that the platform could effectively improve the robot image recognition cycle and offload the computing load.

A Back Propagation Neural Network Sliding Mode Controller for Ship Course Nonlinear System

A back propagation network sliding mode variable structure control method is proposed for ship course nonlinear control. Neural network is used to simulate the functional relation between state hyperplane of the system and the exponential reaching law. A hyperbolic tangent function is applied to replace the saturation function to realize the boundary method of sliding mode control. Chattering is greatly reduced and simulation results demonstrate the presented control method has good adaptability and high robustness.