Sensorimotor learning for artificial body perception
German Diez-Valencia, Takuya Ohashi, Pablo Lanillos*, Gordon Cheng This work was supported by SELFCEPTION project (www.selfception.eu) European Union Horizon 2020 Programme (MSCA-IF-2016) under grant agreement no. 741941. Workshop on Crossmodal Learning for Intelligent Robotics. IEEE Int. Conference on Intelligent Robots and Systems (IROS 2018) Department of Electrical Engineering and Computer Sciences Technical University of Munich Munich, Germany
Abstract
Artificial self-perception is the machine ability to perceive its own body, i.e., the mastery of modal and intermodal contingencies of performing an action with a specific sensors/actuators body configuration [1]. In other words, the spatio-temporal patterns that relate its sensors (e.g. visual, proprioceptive, tactile, etc.), its actions and its body latent variables are responsible of the distinction between its own body and the rest of the world. This paper describes some of the latest approaches for modelling artificial body self-perception: from Bayesian estimation to deep learning. Results show the potential of these free-model unsupervised or semi-supervised crossmodal/intermodal learning approaches. However, there are still challenges that should be overcome before we achieve artificial multisensory body perception.
中文速览
让机器人像人一样感知自己的身体——分清哪些是"自己"、哪些是"外部世界"——是机器人自主性和安全交互的核心难题。研究者依次尝试了三类方法：用层次贝叶斯模型融合视觉与加速度计信息来判断图像区域是否属于机器人自身；用基于预测编码的生物可信模型估计身体潜变量，并在机器人上成功复现了"橡皮手错觉"中的本体感觉漂移现象；以及用生成对抗网络（GAN）学习关节角度与视觉外观之间的前向模型，并进一步实现视觉、触觉、本体感觉三种模态之间的跨模态信号重建。实验结果表明，这些无监督或半监督方法已能在真实或仿真机器人上初步实现多感官身体自感知，但跨模态架构如何兼顾不同传感器的数据特性仍是有待突破的关键挑战，解决这一问题将直接提升机器人的环境适应性与人机交互安全性。
原文 arXiv:1901.09792;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1901.09792v1