Uncertainty has been an important topic, in research, as well as a social concern. The notion of path uncertainty is introduced as the likelihood of encountering a wide variety of possible trajectories when following a given strategy. The research question is: “How can path uncertainty be modelled?”. This thesis proposes the Path Uncertainty Aware Markov Decision Process (PUA-MDP), based on other types of MDPs related to other types of uncertainty. Its algorithm finds optimal policies for balancing maximal reward with minimal cumulative path uncertainty exposure. Experimental validation demonstrates that the algorithm’s behaviour resembles human behavioural responses to uncertainty. It also demonstrates that a small decrease in reward can result in a drastic decrease in uncertainty. If such a method is applied to any classic MDP, path uncertainty could be reduced greatly.
Identifer | oai:union.ndltd.org:UPSALLA1/oai:DiVA.org:umu-227407 |
Date | January 2024 |
Creators | de Graaf, Anaïs |
Publisher | Umeå universitet, Institutionen för datavetenskap |
Source Sets | DiVA Archive at Upsalla University |
Language | English |
Detected Language | English |
Type | Student thesis, info:eu-repo/semantics/bachelorThesis, text |
Format | application/pdf |
Rights | info:eu-repo/semantics/openAccess |
Relation | UMNAD ; 1500 |
Page generated in 0.0018 seconds