弯曲起伏的采煤工作面两道运人研究与应用

针对材料道超长、巷道条件复杂的大型综采工作面的人员进出困难问题,综合各种设备性能,研制了一种适合弯曲起伏巷道的运送人员装置,从而缓解了人员进出距离远、劳动强度大的问题,创新了工作面的运人方式。

青岗坪煤矿辅助提升及运输系统改造

针对青岗坪矿井现有副斜井串车提升运员、采区巷道运输能力偏小等问题,对该矿辅助提升及运输系统进行改造设计研究。从安全角度出发,分别提出了主斜井架空乘人器方案、副斜井异形轨卡轨人车方案;一采区上山辅助运输提出了单轨吊运输方案、异形轨齿轨卡轨车运输方案、无极绳连续牵引车分段接力运输3个方案。该设计改造具有一定的推广应用价值。

Wavelet network solution for the inverse kinematics problem in robotic manipulator

Wavelet network, a class of neural network consisting of wavelets, is proposed to solve the inverse kinematics problem in robotic manipulator. A wavelet network suitable for dealing with multi-input and multi-output system is constructed. The network is optimized by reducing the number of wavelets handling large dimension problem according to the sample data. The algorithms for sparseness analysis of input data and fitting wavelets to the output data with orthogonal method are introduced. Then Levenberg-Marquardt algorithm is used to train the network. Simulation results showed that this method is capable of solving the inverse kinematics problem for PUMA560.

HIT anthropomorphic robotic hand and finger motion control

Nowadays many anthropomorphic robotic hands have been put forward. These hands emphasize different aspects according to their applications. HIT Anthropomorphic Robotic Hand (ARhand) is a simple, lightweight and dexterous design per the requirements of anthropomorphic robots. Underactuated self-adaptive theory is adopted to decrease the number of motors and weight. The fingers of HIT ARhand with multi phalanges have the same size as those of an adult hand. Force control is realized with the position sensor, joint torque sensor and fingertip torque sensor. From the 3D model, the whole hand, with the low power consumption DSP control board integrated in it, will weigh only 500 g. It will be assembled on a BIT-Anthropomorphic Robot.

Multimodal alarm behavior in urban and rural gray squirrels studied by means of observation and a mechanical robot

Urbanization of animal habitats has the potential to affect the natural communication systems of any species able to survive in the changed environment. Urban animals such as squirrels use multiple signal channels to communicate, but it is un- known how ttrbanization has affected these behaviors. Multimodal commtmication, involving more than one sensory modality, can be studied by use of biomimetic mechanical animal models that are designed to simulate the multimodal signals and be pre- sented to animal subjects in the field. In this way the responses to the various signal components can be compared and contrasted to determine whether the multimodal signal is made up of redundant or nonredundant components. In this study, we presented wild gray squirrels in relatively urban and relatively rural habitats in Western Massachusetts with a biomimetic squirrel model that produced tail flags and alarm barks in a variety of combinations. We found that the squirrels responded to each unimodal component on its own, the bark and tail flag, but they responded most to the complete multimodal signal, containing both the acoustic and the moving visual components, providing evidence that in this context the signal components are redundant and that their combination elicits multimodal enhancement. We expanded on the results of Partan et al. （2009） by providing data on sig- naling behavior in the presence and absence of conspecifics, suggesting that alarm signaling is more likely if conspecifics are present. We found that the squirrels were more active in the urban habitats and that they responded more to tail flagging in the urban habitats as compared to the rural ones, suggesting the interesting possibility of a multimodal shift from reliance on audio to visual signals in noisier more crowded urban habitats [Current Zoology 56 （3）： 313-326, 2010].

A new accelerating algorithm for multi-agent reinforcement learning

In multi-agent systems, joint-action must be employed to achieve cooperation because the evaluation of the behavior of an agent often depends on the other agents’ behaviors. However, joint-action reinforcement learning algorithms suffer the slow convergence rate because of the enormous learning space produced by joint-action. In this article, a prediction-based reinforcement learning algorithm is presented for multi-agent cooperation tasks, which demands all agents to learn predicting the probabilities of actions that other agents may execute. A multi-robot cooperation experiment is run to test the efficacy of the new algorithm, and the experiment results show that the new algorithm can achieve the cooperation policy much faster than the primitive reinforcement learning algorithm.

Artificial molecular machines that can perform work

An artificial molecular machine consists of molecule or substituent components jointed together in a specific way so that their mutual displacements could be initiated using appropriate outside stimuli. Such an ability of performing mechanical motions by consuming external energy has endowed these tiny machines with vast fascinating potential applications in areas such as actuators, manipulating atoms/molecules, drug delivery, molecular electronic devices, etc. To date, although vast kinds of molecular machine archetypes have been synthesized in facile ways, they are inclined to be defined as switches but not true machines in most cases because no useful work has been done during a working cycle. More efforts need to be devoted on the utilization and amplification of the nanoscale mechanical motions among synthetic molecular machines to accomplish useful tasks. Here we highlight some of the recent advances relating to molecular machines that can perform work on different length scales, ranging from microscopic levels to macroscopic ones.