Toward an idiomatic framework for cognitive robotics

ABSTRACT: Summary Inspired by the “cognitive hourglass” model presented by the researchers behind the cognitive architecture called Sigma, we propose a framework for developing cognitive architectures for cognitive robotics. The main purpose of the proposed framework is to ease development of cognitive architectures by encouraging cooperation and re-use of existing results. This is done by proposing a framework dividing development of cognitive architectures into a series of layers that can be considered partly in isolation, some of which directly relate to other research fields. Finally, we introduce and review some topics essential for the proposed framework. We also outline a set of applications. Graphical abstract Highlights • The proposed framework divides development of cognitive architectures into layers• The framework spans the best-known approaches employed within cognitive robotics• The framework is centered around modern probabilistic programming techniques• Two applications demonstrate the concepts of the framework The bigger picture For many decades, robots have been expected to transfigure the world we live in, and in many ways they already have, by increasingly taking over dull, dirty, and dangerous jobs. However, for robots to integrate fully and seamlessly into human societies, robots need to be able to learn and reason from experience and effectively deal with unpredictable and dynamic environments. Developing robotic systems with such intelligence is a tremendous and difficult task, which has led to the foundation of the new multi-disciplinary scientific field called cognitive robotics, merging research in adaptive robotics, cognitive science, and artificial intelligence. To ease merging research from these scientific fields, we propose a general framework for developing intelligent robotic systems based on recent advancements in the machine learning community. We hope that this framework will aid researchers and practitioners in bringing even more helpful robots into our societies. Damgaard et al. propose a general framework for developing cognitive architectures for robots. The proposed framework divides development into a series of layers separating models of cognition from the algorithms that implement them. Based on modern probabilistic programming and variational inference algorithms, the proposed framework aims to support the four desiderata generic cognition, grand unification, functional elegance, and sufficient efficiency. The concepts of the framework are demonstrated through two robotic applications.

SUBMITTER: Damgaard M 

PROVIDER: S-EPMC9278519 | biostudies-literature |

REPOSITORIES: biostudies-literature