【摘要】 机器人抓取模式选择主要是利用人的抓取经验来进行的,具有一定的不确定性和模糊性。本文根据这一特点,以研制的形状自适应手爪抓取模式分类为基础,在综合考虑抓取任务和物体特征的同时。采用模糊的输入方式,同时在保持分类能力不变的前提下,采用粗糙集理论从训练样本中提取和精简规则来构建模糊神经网络。利用神经网络良好的分类特性来选择合适的抓取模式,减少了网络输入,简化了网络拓扑结构,缩短了训练时间,提高了抓取的自动化水平。最后通过抓取实验验证了抓取模式选择的正确性。更多还原

【Abstract】 The choice of robot hands grasp mode is usually achieved based on human grasp experience, thus is characterized by fuzziness and uncertainty to some extent. According to these characteristics, by employing a fuzzy neural network, this paper has established a proper map between the fuzzy inputs, including grasp task and object features, and the outputs, namely the versatile grasp modes of the shape self--adapting robot hand. On the primes of keeping the classification ability of the neural network, the rough set theory is used to deduce and simplify the grasp rules. The established fuzzy neural network has a better topological structure with greatly reduced network scale, less network inputs and improved training speed. At last grasp experiments validate the correctness of the grasp mode choice.更多还原