论文部分内容阅读
信物融合在机器人应用中具有重要作用,其关键技术支撑是多模态信息融合技术。该文分析了机器人灵巧手系统中广泛应用的触觉模态和深度模态的特点,采用隐层空间的思想建立了单一模态的动态模型;针对传统Bayes融合在机器人应用中出现的联合似然易受到不准确似然影响的问题,根据单一模态动态模型的特点,提出了一种具有信任度评价的Bayes多模融合策略。它可以根据当前结果更新信任度评价函数,修正原始不准确似然,动态改善融合策略的适应性和准确性。实验结果表明:该方法可以克服单一模态决策的不足,有效利用多模异构信息的优点进行可靠的融合决策,分类正确率较传统Bayes融合有显著提高。
Trident fusion plays an important role in robotics applications, and its key technical support is multi-modal information fusion technology. This paper analyzes the characteristics of the tactile modes and depth modes that are widely used in robot dexterous hand systems, and adopts the idea of hidden layer space to establish a single modal dynamic model. Aiming at the joint likelihood of traditional Bayesian fusion in robot applications, According to the characteristics of a single modal dynamic model, a Bayes multimode fusion strategy with trust evaluation is proposed. It can update the trust evaluation function according to the current results, correct the original inaccurate likelihood and dynamically improve the adaptability and accuracy of the fusion strategy. The experimental results show that the proposed method can overcome the shortcomings of single modal decision making and make effective fusion decision by using the advantages of multi-mode heterogeneous information effectively. The classification accuracy is significantly improved compared with traditional Bayesian fusion.