Master Thesis – 2019
DEVELOPING AN ETHICS MODULE FOR A SERVICE ROBOT
With the improvements of robotics science, robots have gradually began to take their place in social environments. The human-robot interaction (HRI) studies evolved with the increasing number of robots involved in the social world. It is impossible to predict whether this changed relationship will be competent or corrupt. The most significant anxiety of humanity regarding robots is that one or a group of robots dominate the world and create an apocalypse for humankind. The science of robot ethics has emerged to prevent this possible disaster scenario and to define the limits of HRI.
There are many approaches in the literature about how robot ethics should be and what to expect from robot ethics. In this thesis, I examined the applied ethics approaches and designed an ethical unit for the service robot BOSS, which was developed by our lab. Our ethics unit works as a expert system using fuzzy logic, which is called Fuzzy Expert System (FES) in machine learning. The purpose of designed FES is to enable the robot to approach the human being more ethically than any person. There are two kinds of ethics rules in our FES. One is a long-term memory of ethics that is universal consent and set in stone, and the other one is a short-term memory of ethics rules that will alternate according to the working environment and duty of the robot.
The ethical module takes the possible behavior of the robot from the behavior controller and the environmental perception as inputs. By combining these inputs with rules created from fuzzy clusters, the robot will choose the most ethical, so the most harmless to the user, of the possible behaviors. If there is no ethical behavior, the robot will stop acting and show no action. In the first stage of research, we determined the outline of the ethics module and ethical parameters according to the literature.
Bachelor thesis – 2015
Development of Coach Robot for Robot Soccer Team
In the robot soccer, understanding the environment ,namely the world model, appropriately and displaying the behaviors according to the world model are crucial. In Standard Platform League, robots can individually adapt dynamic environment comfortably thanks to advanced localization and cognition techniques but adapting as a team is a challenging problem. In order to solve this problem, “Coach Robot” concept is introduced in SPL in this year.
This work proposes a coach robot design and implementation with decision tree structure and Kalman filter detection system. The constructed world model is divided into a set of layers and for every layer a set of new concepts (roles, formations, set plays) are defined in order to understand the environment more appropriate. A set of solutions (strategies) and a range of parameters are determined according to newly defined concepts. We represent that our “Coach Robot” approach has advantages over the previous system in terms of forming team strategy and dynamic strategy change for adapting to the environment.