Explore chapters and articles related to this topic
Bioelectrical coordination of cell activity toward anatomical target states
Published in David M. Gardiner, Regenerative Engineering and Developmental Biology, 2017
Celia Herrera-Rincon, Justin Guay, Michael Levin
We define goal-directed decision-making process as the ability of a system to (1) infer the future state as a function of both current and past states (self and environment), and (2) regulate its behavior, in a flexible manner, for orienting it to the goal (Verschure et al. 2003, 2014, Pezzulo et al. 2014, 2015). In engineering terms, this is an optimization or minimization (least-action) process, which seeks to reduce some quantity (e.g., the difference between current, damaged, anatomical state and the target morphology) (Kaila and Annila 2008, Friston et al. 2012, Friston et al. 2015). The decision-making process integrates four different cognitive modules: perception and attention (selection of relevant information—self-state and external world and goals), memory (mostly recall), prediction and valuation (reward), and selection and monitoring (planning of active behavior, such as motor sequences, and performance monitoring, with subsequent error detection and adjustment).
Special issue on symbol emergence in robotics and cognitive systems (II)
Published in Advanced Robotics, 2022
Tadahiro Taniguchi, Takayuki Nagai, Shingo Shimoda, Angelo Cangelosi, Yiannis Demiris, Yutaka Matsuo, Kenji Doya, Tetsuya Ogata, Lorenzo Jamone, Yukie Nagai, Emre Ugur, Daichi Mochihashi, Yuuya Unno, Kazuo Okanoya, Takashi Hashimoto
Innovations in artificial intelligence have opened the door to the next generation of cognitive robotics. Indeed, deep learning and statistical machine learning provide a wide range of cognitive modules, e.g. image and speech recognition, localization and mapping, natural language processing, and motion planning. However, most of the achievements in artificial intelligence are made in computer and simulation environments. In contrast with the conventional artificial intelligence that is optimized with large amounts of prepared data, we, human beings, learn a variety of skills and knowledge from our own sensorimotor experiences. To develop a cognitive and developmental robot that can learn and adapt in real environments, we still have a huge number of challenges because the world is full of uncertainty and dynamic changes. Also, emphasizing the developmental aspects of cognition is important to understand the developmental process of human cognitive systems.
Special issue on Symbol Emergence in Robotics and Cognitive Systems (I)
Published in Advanced Robotics, 2022
Tadahiro Taniguchi, Takayuki Nagai, Shingo Shimoda, Angelo Cangelosi, Yiannis Demiris, Yutaka Matsuo, Kenji Doya, Tetsuya Ogata, Lorenzo Jamone, Yukie Nagai, Emre Ugur, Daichi Mochihashi, Yuuya Unno, Kazuo Okanoya, Takashi Hashimoto
Innovations in artificial intelligence have opened the door to the next generation of cognitive robotics. Indeed, deep learning and statistical machine learning provide a wide range of cognitive modules, e.g. image and speech recognition, localization and mapping, natural language processing, and motion planning. However, most of the achievements in artificial intelligence are made in computer and simulation environments. In contrast with the conventional artificial intelligence that is optimized with large amounts of prepared data, we, human beings, learn a variety of skills and knowledge from our own sensorimotor experiences. To develop a cognitive and developmental robot that can learn and adapt in real environments, we still have a huge number of challenges because the world is full of uncertainty and dynamic changes. Also, emphasizing the developmental aspects of cognition is important to understand the developmental process of human cognitive systems.