An announcement of the course can also be found in the course information system CAMPUS.
There is an L²P-Lernraum for this course. You need to register for this room by entering the registration on the CAMPUS site above while you are logged in with your CAMPUS-Office account.
Agents are autonomous, computational entities (programs, robots) which perceive their environment through sensors and which are able to take influence on it through effectors. Many entities in robotics such as the environment itself, sensing results, and action effects are uncertain. To be able to cope with those uncertainties in complex domains one usually uses probabilistic mechanism. Many of those mechanism employ the Markov Assumption which refers to the dependence of temporal states. With it, values in any given state are only influenced by the values of the state directly preceding it.
We will cover various methods and formalism using the Markov Assumption like Markov Chains, Markov Decision Processes (in several forms), and Hidden Markov Models and investigate their applications to robotics. Theoretical backgrounds as well as practical aspects will be discussed and illuminated.
Participants have already been selected and notified.
- basic study period completed (Vordiplom)
- profound skills in mathematics, especially probability theory
|Paul Köhler||HMMS for Speech Recognitin||STF|
|Matthias Schiffer||HMMS for Complex Action Recognition||DB|
|Dan Zhou||Markov Chain Monte Carlo Data Association||STF|
|Michael Wolf||Markov Chain Monte Carlo for Topological Maps||DB|
|Patrick Jüptner||EKF-Markov Localization||STF|
|Damian Lukowski||Monte Carlo Localization||DB|
|Deni Raco||EKF Visual Tracking (Condensation Alg.)||STF|
|Christian Bien||Rao-Blackwellised Particle Filter visual SLAM||DB|
|Carsten Cordes||Markov Decision Processes - Planning||STF|
|Daniel Schmitz||Markov Decision Processes - Learning||DB|
|Christian Ferber||Markov Decision Processes - Adv. Topics||STF|
|Chitra Hapsari Ayuningtyas||Semi-MDPs and continuous Time||DB|
Here is the list of literature.
Seminar talks will be held weekly throughout the semester. Participation is bindingly for achieving the credit. The preparation of the talks and the written workout should not be a translation or renarration of the given paper(s). The given literature is meant to be a starting point for own literature investigation on the topic. The literature should then be discussed with the instructor.
No later than 6 weeks before the talk the seminarist should hand in a preliminary workout and 3 week (latest!) before the talk the final version of the slides/transparencies must be given to the respective seminar instructor.
You are of course more than welcome to discuss and coordinate with us way before those final deadlines.
Instructions on a good seminar talk
Renate Eschenbach from our library offers guided tours on how to find literature in the library and how to prepare a seminar.
Interested students should enlist for a tour in the preliminary discussion.