Search

Path planning is one of the key technologies of robot navigation and control, in path planning, there are some problems in the application of ant colony algorithm, such as slow convergence speed, poor optimization results and incomplete search. In order to improve the mobile robot's ability to search the optimal path to the target point in the global static environment. In this paper, a double improved fusion algorithm of particle swarm optimization and ant colony algorithm is proposed to solve the path planning problem. Firstly, the occupied grid map is constructed based on visual slam technology of depth camera in static environment; Secondly, the improved particle swarm optimization -ant colony algorithm is used for path planning in the grid map: The sub optimal solution is obtained by using the advantages of global search ability and search speed of improved particle swarm optimization, which is transformed into the increment of initial pheromone distribution in the improved ant colony algorithm, and the exact solution of the path problem is solved by using the positive feedback mechanism of the improved ant colony algorithm; Finally, a robot experimental platform is built to verify the effectiveness and practicability of the improved particle swarm optimization and ant colony fusion algorithm. The experimental results show that the fusion algorithm has a certain guiding role for mobile robot path planning.

In order to solve the problem of rust removal on the inner and outer surface of petrochemical storage tanks that consumes a lot of manpower and material resources during the operation, a derusting wall climbing robot of permanent magnetic adsorption type was designed. Firstly, the mechanical structure design idea is carried out. Then, the force received by the robot which affects the stability of the robot movement is analyzed by the statics motion calculation formula. Finally, the real-time control system software of the robot is designed to realize the movement of the robot. The purpose of this paper is making the robot to be a remote control tool to help workers finishing petrochemical storage tanks rust removal operation.

In order to reduce the radiation dose of operators and the labor intensity of staff during the nuclear reactor operation, a remote operation nuclear multifunction robot was designed by service for salvage operation in the nuclear reactor pool. The purpose of this paper is making the robot to be a remote control tool to help staff in the nuclear reactor pool. Firstly, the kinematics analysis of the robot's six-wheeled mobile platform and four-degree-of-freedom manipulator are carried out. Then, based on the whole results of kinematics analysis, the overall mechanical structure of the robot is designed radiation-resistant and compact.

Because of the robot’s special application background, such as underwater, nuclear environments, space, and disaster sites, after Fukushima nuclear spill the robot for nuclear power station has become a hot spot in the field of robotics research. A remote operation nuclear multifunction robot was designed by service for salvage operation in the nuclear reactor pool to replace operators’ dangerous and complicated work. In order to meet the overall performance requirements of the robot, during the design phase of the complete machine, the designer selected hardware suitable for the robot’s indicators and designed software that can apply the selected hardware. The grabbing foreign materials experiment was tested in order to verification the control system development satisfies the design indexes. The result shows that the robot can grabbing foreign materials rapidly and exactly.

A novel wideband filtering balun power divider using Schiffman phase shifter is proposed. A modified Wilkinson power divider and two Schiffman phase shifters are adopted to realize wideband balun power divider. In addition, two transmission zeros are realized by using an open-ended stub at the input port to realize the filtering characteristic. A wideband filtering balun power divider ( $\varepsilon_{\mathrm{r}}= 3.66,\ \mathrm{h}=0.762\ \text{mm},\ tan\delta=0.004$ ) centering at 2 GHz is designed and fabricated for verification.

Linear prediction code (LPC) has been applied not only on speech processing but also loudspeaker equalization. There are shortcomings processing low frequency and overcome by warped frequency. A loudspeaker system has been equalized by LPC and warped-frequency linear prediction (WLPC), three conclusions have been drawn. First, LPC has higher frequency resolution and lower effective frequency with more orders. Secondly, WLPC processes low frequency better than LPC with the same order, and it has some better equalized results with more orders. Thirdly, both methods maybe determine incorrectly the “peak and valley” if orders are not enough and lose optimum equalized results.

A variety of masking schemes have been introduced to protect the block cipher. The critical technology lies on how to implement the nonlinear part with minimal resources. One popular way to solve this problem is using tower field representation. However, this idea always need two variables as inputs for conversion, which may result in suspicious first-order leakages during this operation. In this paper, we apply first-order attacks against variables of the conversion part based on simulation experiments to prove our speculation. The simulation results show that our proposed idea is more efficient than the second-order attacks in the noisy scenario.

Higher Order Differential Power Analysis (HO-DPA) has been introduced for more than ten years. Various combining functions are applied to attack the masked AES implementation, but finding an appropriate combining function is still a big issue. We study the weakness of absolute difference method when using multi-bit variables to mount an attack. In this paper, an optimized method is proposed to improve its efficiency. Simulation results based on different combining functions show that the success rate of our method is twice as high as that of normal absolute difference and sum combining function. Our idea is also confirmed by physical traces measured on a FPGA platform.

MANET mesh-based multicast routing protocols cost much more overhead because of redundant packet forwarding. Considering limited energy of wireless nodes, this paper presents an energy efficient multicast routing tree algorithm, which had the least forwarding nodes. The algorithm was proved to be a minimum cost routing tree algorithm in wireless network. Also, a practical distributed version of the algorithm was presented. Simulation comparisons show that improved ODMRP by the distributed algorithm has a distinct decrease in transmission number, which is close relative to energy consumption. It is especially obvious when there are more multicast receivers in MANET.

AdaBoost has been proved to be an effective method to improve the performance of base classifiers both theoretically and empirically. However, previous studies have shown that AdaBoost cannot obviously improve the performance of Naive Bayes as expected. This paper presents a new boosting algorithm, ActiveBoost, which applies active learning to mitigate the negative effect of noise data and introduce instability into boosting procedure. Empirical studies on a set of natural domains show that ActiveBoost has clear advantages with respect to the increasing of the classification accuracy of Naive Bayes when compared against AdaBoost.

This paper describes a new data mining algorithm to learn Bayesian networks structures based on an extending evolutionary programming (EP) method and the minimum description length (MDL) principle. Aiming at preventing and overcoming the premature convergence, the algorithm combines the niche technology into the selection mechanism of EP. In addition, our algorithm, like some previous work, does not need to have a complete variable ordering as input. To evaluate the performance of our algorithm, we conduct a series of experiments and compare them with the previous work based on genetic algorithms (GA). The experimental results illustrate that both the quality of the solutions and computational time of our algorithm are superior.

This paper describes a new data mining algorithm to learn Bayesian networks structures from incomplete data based on an extending evolutionary programming (EP) method and the minimum description length (MDL) principle. This problem is characterized by a huge solution space with a highly multimodal landscape. The algorithm presents fitness function based on expectation, which converts incomplete data to complete data utilizing current best structure of evolutionary process. Aiming at preventing and overcoming premature convergence, the algorithm combines the niche technology into the selection mechanism of EP. In addition, our algorithm, like some previous work, does not need to have a complete variable ordering as input. The experimental results illustrate that our algorithm can learn a good structure from incomplete data.