Skip to main content

Master Thesis Proposals

Last updated: 2021 September

Automatic Fault Detection and Recovery for Autonomous Underwater Vehicles (AUVs) using Behavior Trees

Start date: Anytime

Contact: Özer Özkahraman,

A Behaviour Tree (BT) is a decision making method that makes use of a tree structure and specific well-defined nodes in the tree in order to control a system to a desired state[1,2]. BTs are currently being used in AUVs to control the mission flow and autonomous decision making.

AUVs are expensive robots that can spend very long times without any possibility of human contact or monitoring, in hostile and unknown environments. Often times they are expected to explore these remote areas with absolutely no supervision and the data they acquire is one-of-a-kind. Due to these harsh conditions and high expectations, a failure during a mission can be disastrous.

The interested student is expected to look into methods for autonomously detecting failures on AUVs during a mission, such as a propleller getting stuck, losing a fin and leaks. The AUVs can be very maneuverable and thus they can attempt to maneuver into specific states to experiment and ascertain the damage. If this part is accurate enough, the next step would be to find the states that the vehicle can still make use of. For example if the AUV has 2 propellers and one of them is only working at 50%, the AUV can still go forward if it uses the undamaged prop at 50% too and use this to make it back home.


Automatic Fault Detection and Recovery with MULTIPLE Autonomous Underwater Vehicles (AUVs)

Start date: Anytime

Contact: Özer Özkahraman,

Same problem and setup as the above, but this time instead of the AUV detecting its own fault, it can call for another AUV to help get an "outside view". Usually AUVs are equipped with forward looking sonars that can detect and measure another AUV, together with low-bandwidth (10s of bits per second) communication means. Which can allow, at best, movement-based detection.


Rendezvous Guarantees for Collaborative Coverage for AUVs

Start date: Anytime

Contact: Özer Özkahraman,

Using multiple small AUVs that collaboratively do coverage and help each other with localization improves efficiency and increases the total possible coverage with higher accuracy. All these nice properties hinge on the AUVs being able to rendezvous with each other to exchange measurements and other data. Unfortunately, underwater there are currents and unknown dynamics in effect, meaning the AUVs will drift. This causes them to miss their rendezvous times and locations, leading to the entire system to lose all of its nice properties.

The interested student is expected to look into methods on how rendezvous can be guaranteed under such harsh conditions. It is expected that there will be a trade-off between rendezvous probability and efficiency, finding the relationship between them is also very valuable. There are multiple ways to approach this problem, model-predictive control, RRT* based approaches, conservative limits based approaches and so on.


[1] Colledanchise, Michele, and Petter Ögren. Behavior Trees in Robotics and Al: An Introduction. CRC Press, 2018.
[2] Colledanchise, Michele, and Petter Ögren. "How behavior trees modularize robustness and safety in hybrid systems." 2014 IEEE/RSJ International Conference on Intelligent Robots and Systems. IEEE, 2014.

Profile picture of Özer Özkahraman


  • Master Thesis Proposals