My PhD Thesis: "Decision-Making under Uncertainty for Real Robot Teams"


On February 26th, 2014, I successfully defended my PhD thesis at IST. My thesis was awarded the grade of "Pass with Distinction", the highest of the possible grades at IST.

You can find the full manuscript of my thesis here.

pic1 pic2

Supporting Material

In the following links, you can find videos containing experimental results related to my PhD thesis, and the various problem files that were used to implement my decision-theoretic models in practice.


Chapter 3 - Robotic Soccer Discrete MPOMDP
[Video 1]: Unrestricted dribbling behavior (Fig. 3.9 of the thesis).
[Video 2]: Passing behavior (Figure 3.10 of the thesis).
[Video 3]: Failed passing behavior due to loss of reactivity.
Chapter 5 - Robotic Soccer GSMDP/MMDP
[Video]: A comparison of the synchronous MMDP and event-driven GSMDP implementations, both in simulation and with the real robots.
Chapter 7 - Multi-Robot Surveillance Case Study
[Video]: The implementation details and experimental results of our autonomous surveillance system.

Problem Files

Chapter 3
[Robotic Soccer Discrete MPOMDP (.dpomdp)]
(Note: .dpomdp files can be used with the MADP Toolbox)
Chapter 4
[Relay-Small (.dpomdp)]
[Relay-Large (.dpomdp)]
(Note: The OneDoor problem referred in the results of Chapter 4 can be found in the MADP problem repository.)
Chapter 5
[Robotic Soccer MMDP Model (.mat)]
[Robotic Soccer GSMDP Model (.mat)]
(Note: These are MATLAB files containing transition and reward matrices, intended to be used with the MDP Toolbox for MATLAB (Not to be confused with the MADP toolbox!))
Chapter 6
[Access2 - Event Driven (.pgmx)]
[Access2 - Synchronous (.pgmx)]
[Access3 - Event Driven (.pgmx)]
[Access3 - Synchronous (.pgmx)]
(Note: .pgmx files can be used with the OpenMarkov graphical model software.)
Chapter 7
[Coordinative Event Driven MPOMDP (.pgmx)]
[Patrol Event Driven POMDP (.pgmx)]
[Assistance Response FSM (.py)]
[Surveillance Incident Response FSM (.py)]
[Emergency Response FSM (.py)]
(Note: the Python scripts contain SMACH finite-state machines.)