Start my eKVV Studieninformation Lernräume Prüfungsverwaltung Bewerbungs-/Statusportal Anmelden

392162 Learning dynamic behavior generation from human-like mental states (Pj) (SoSe 2019)

Inhalt, Kommentar

Human behavior is governed by their mental state. Desires dictate goals
and different beliefs will
influence planning strongly.

Suppose an agent tries to navigate a partially known environment. It may
know the general layout of
routes, but is not entirely sure about the presence or absence of
obstacles along some paths. If the
agent does not know about an obstacle, it will most likely plan behavior
resulting in it encountering
said obstacle before having to replan in order to finally reach its
desired goal.
Conversely, if the agent incorrectly assumes the presence of an obstacle
on the shortest
path, the agent might take an unnecessary detour to avoid it.

The goal of this project is to train (e.g. using Reinforcement Learning)
such an agent which will be able
to generate such behavior given certain mental states and an environment
while dynamically reacting to changes
in the (assumed) environment.


Termine (Kalendersicht )

Rhythmus Tag Uhrzeit Ort Zeitraum  

Zeige vergangene Termine >>


Modul Veranstaltung Leistungen  
39-M-Inf-P Projekt Projekt unbenotete Prüfungsleistung
39-M-Inf-P1_NWI Projekt 1 Projekt 1 unbenotete Prüfungsleistung

Die verbindlichen Modulbeschreibungen enthalten weitere Informationen, auch zu den "Leistungen" und ihren Anforderungen. Sind mehrere "Leistungsformen" möglich, entscheiden die jeweiligen Lehrenden darüber.

Konkretisierung der Anforderungen
Keine Konkretisierungen vorhanden
Automatischer E-Mailverteiler der Veranstaltung
Änderungen/Aktualität der Veranstaltungsdaten