China
Africa
Explore chapters and articles related to this topic
Humans, Machines, and Social Cognition
Published in Alessio Plebe, Pietro Perconti, The Future of the Artificial Mind, 2021
Alessio Plebe, Pietro Perconti
It is interesting to note that the attributive mechanisms mentioned above, i.e., anthropomorphic mental triggers, are good guides for designing both humanoid bodies equipped with the right mentalization cues and human robot heads equipped with the same cognitive abilities to detect these cues in overt human behavior. Virtual agents can mimic human eye abilities with greater accuracy than embodied robots (Ruhland et al., 2015). Things look a little worse for humanoid robots. Even the best robots are not yet capable of pupil dilation, although this behavior is an indicator of the mental state of people (Hyönä et al., 1995; Hanson and White, 2004; Admoni and Scassellati, 2017; Chevalier et al., 2020). Among the (anthropomorphic) mental triggers is the ability to detect the movements performed by biological agents in the environment (Johansson, 1973). It seems that human infants of less of 3 months of age are endowed with this capacity (Simion et al., 2008). The same ability should be given to humanoid robots if we are interested in an ecological Human-Robot Interaction (Vignolo et al., 2017). The idea is to equip robots with a module integrated in the software as, for example, in the iCub humanoid robot software (Vignolo et al., 2016). The iCub robot is designed as a baby humanoid, as visible in Figure 6.3, and is aimed at developmental robotics research (Cangelosi and Schlesinger, 2015).
A Panoramic Survey on Grasping Research Trends and Topics
Published in Cybernetics and Systems, 2019
Manuel Graña, Marcos Alonso, Alberto Izaguirre
Almost all brands of computational intelligence techniques have been tested for grasping sensing and actuation. Interpretation of tactile information is one of the hot computational intelligence challenges. Compressed learning has been demonstrated in the classification of tactile signals from arrays of tactile sensors embedded in a robotic skin working on the natural low dimension data obtained directly from the sensors (Hollis, Patterson, and Trinkle 2018). Another recent proposal encodes the signals using a linear dynamic system approach applying a fuzzy c-means clustering algorithm to the signal features, the resulting signal encodings are used for recognition in a bag-of-words approach (Liu et al. 2018). Combining particle filters and unscented Kalman filter (Vezzani et al. 2017) are able to overcome the strong non-linearities and multimodal distributions of tactile measurements in time achieving localization of objects in 6 degrees of freedom and accurate geometrical modeling with a commercial iCub robot endowed with capacitive fingertip sensors. Random forests have been used to classify tactile information predicting slippage in order to achieve grasp stabilization in a collaborative experiment with a human and robot endowed with a novel tactile sensor (Veiga, Peters, and Hermans 2018). The publication of rich datasets, such as the so-called YCB dataset (Calli et al. 2017), has fostered research from many fronts that serve both to validate published approaches and to try new approximations to grasping control issues.
Probabilistic movement primitives under unknown system dynamics
Published in Advanced Robotics, 2018
Alexandros Paraschos, Elmar Rueckert, Jan Peters, Gerhard Neumann
Further, we perform evaluations on complex real-robot platforms. The humanoid robot iCub lifts a grate to a predetermined height from different grasping locations, without learning a model of the grate. Subsequently, we evaluate our approach on moving a chemistry flask of an unknown weight to a target location, while avoiding obstacles, using the KUKA LWR robotic arm.