University of Twente Student Theses


Human perception of an adaptive agent's fear simulated based on TDRL Theory of emotions

Dai, L. (2019) Human perception of an adaptive agent's fear simulated based on TDRL Theory of emotions.

[img] PDF
Abstract:To better collaborate with humans, robots need to be designed to learn from humans and the environment to function autonomously. Reinforcement Learning(RL) has been used in different areas with successful results in task learning. But the learning process in RL could be hard for humans to understand and therefore not able to give proper feedback to the robot about its behaviour, which is important for autonomous learning. One way to overcome this problem is to add emotions to the robot, as emotions are used by humans to express their internal states. Temporal Difference Reinforcement Learning (TDRL) Theory of Emotion proposes a structure for agents to express appropriate emotions during the learning process. Simulations have been done to test simulate emotions in several scenarios, but there is no further experiment to test how plausible these simulated emotions are when perceived by humans. This thesis aims to find out the plausibility of simulated fear perceived by humans. 237 human participants were recruited to evaluate different fear calculation methods. Results suggest the fear calculation method with e-greedy fear policy(e=0.1) and long-horizon provides a plausible fear estimation, and humans could understand simulated fear based on TDRL Theory of emotions when properly expressed.
Item Type:Essay (Master)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:54 computer science
Programme:Computer Science MSc (60300)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page