27 April 2020
09:00 Master's Defense Fully distance
Theme
Exploring Cognitive Functions in Robotics
Student
Letícia Mara Berto
Advisor / Teacher
Esther Luna Colombini
Brief summary
The advancement of artificial intelligence has brought many benefits to robotics. Today, it is possible to develop robots that not only execute what they were pre-programmed to do, but that learn according to the interaction with the environment and other agents. For this, robots must have cognitive functions, such as memory, decision making, learning, attention, planning and others supported in their structure. To date, there are no standard ways in the literature to assess cognitive architectures. In this context, in this work, we study cognitive functions in order to identify which components are necessary to validate a project that implements a cognitive architecture. From these functions, we designed a set of experiments to determine the robot's behavior and the impact of each module on obtaining knowledge, in addition to defining assessment scenarios. For the development of this research, the CONAIM (conscious attention-based integrated model), a formal model for machine awareness based on an attentional scheme for the cognition of human-like agents and the CST (Cognitive Systems Toolkit), a General tools for building cognitive architectures were used. We improved the bottom-up attentional module of CONAIM, modeled at CST, and an intelligent codelet using Reinforcement Learning was implemented. The tests were carried out in simulation and we were able to successfully control the Pioneer P3-DX robot, learning from its attentional space.
Examination Board
Headlines:
| Esther Luna Colombini | IC / UNICAMP |
| Roseli Aparecida Francelin Romero | ICMC / USP |
| Ricardo Ribeiro Gudwin | FEEC / UNICAMP |
Substitutes:
| Paula Dornhofer Paro Costa | FEEC / UNICAMP |
| Alexandre da Silva Simões | ICTS / UNESP |