John Folkesson

Are you a student at KTH interested in doing research in our AUV group? Read this:   http://www.csc.kth.se/~johnf/Projects_2022.pdf more about SMaRC: SMaRC. Also if you are interested in a PhD in underwater robotics in my group: Underwater Robotics PhD.

My research i in the area of mobile robotics.  The have comprised be underwater, flying, outdoor ground robots, autonomous driving and indoor robots.  Currently the focus in our group  is underwater robots.  The main challenge remains not getting lost, (so called simultaneous localization and mapping SLAM). Place recognition underwater is very different than in man made environments.  Most of the features one observes are dynamic on medium time scales,  ice melts, sea creatures crawl and so on.  Place recognition is intrinsically ambiguous and calls for careful inference.  For autonomous driving I am interested in situation awareness for decision support.  Outdoor robots and indoor robots this has moved into higher levels or spatial representation, semantic mapping, spatial-temporal models.

Underwater SLAM has so far has consistent in relatively few trials of tried and true methods from outdoor land robots to autonomous underwater vehicles, AUV.  This works when only to the extend that the sensors and the environment can be treated as similar to the land analog.  Natural underwater environments are however for the most part quite different.  The features available as navigation aids are not always possible to characterize by the methods common to describing keypoints in images.  Sonar has a much different type of viewpoint dependance as compared to cameras and lidar.  There will always be long transits where the bottom is not in sensor range or where the sensors return 'blank' measurements.  Motion estimation without acoustic contact with the seabed is inherently very uncertain.     New approaches for underwater SLAM will have an impact here.

At the same time AUVs to be effective will need to travel beyond limits of underwater communications.  This makes the future AUV scenarios the most autonomous of all robot scenarios.  These robot must operate without human instruction in areas and situations  that are unknowable at the time of launching the AUV.  Missions under ice  or to the deep ocean are examples.  These envisioned missions will take days to months to carry out with no communications.   Can these robots navigate using the sonar and other clues.  The data that they collect is of little value if it is not geo-referenced.

My home page is here: http://www.csc.kth.se/~johnf/The RPL Division's Master thesis project proposals are here:  RPL Master Theses

Courses:

Probabilistic Graphical Models (DD2420), course responsible, teacher, examiner | Kurswebb

Applied Estimation (EL2320/EL3320), course responsible, teacher, examiner | Kurswebb

The Sustainable Systems and Control Engineer (EL2220), teacher | Kurswebb