Martin Asenov

PhD Candidate

The University of Edinburgh


I am a PhD candidate at The University of Edinburgh and Heriot-Watt University, part of the RAD group, supervised by Dr. Subramanian Ramamoorthy and Dr. Kartic Subr. I am working on the intersection of Robotics, Machine Learning and Physics, i.e. how can we solve highly dynamical tasks in uncertain environments by incorporating physical inductive bias. This allows for performing tasks such as localizing a gas leakage with a UAV, stopping an unknown bouncing ball mid-air, etc., by incorporating useful bias for wind and gas dispersion; free fall and restitution; etc. I am also a committee member of EdIntelligence and cohort representative for the Robotics RAS CDT.


  • Active Sensing
  • Bayesian Optimization
  • Machine Learning
  • Robotics
  • Intuitive Physics
  • Computer Graphics


  • PhD in Robotics and Autonomous Systems

    The University of Edinburgh and Heriot-Watt University

  • MRes in Robotics and Autonomous Systems, 2017

    The University of Edinburgh and Heriot-Watt University

  • BSc in Artificial Intelligence and Computer Science, 2016

    The University of Edinburgh

  • Exchange Student, 2015

    The University of California, Irvine



alternative text for search enginesApplied Scientist Intern

Amazon PrimeAir

Oct 2018 – Dec 2018 Graz, Austria
Worked on image segmentation with domain adaptation.

alternative text for search enginesResearch Scientist


Oct 2017 – Oct 2018 Edinburgh, UK
Worked on gas localization and mapping with a UAV.

alternative text for search enginesSoftware Integration and Characterization Intern


Jun 2015 – Aug 2015 California, United States
Extended and modified Google’s Blockly open source project in order create different scenarios for testing our in-flight entertainment systems. Used nodejs for the server side, thus having almost 100% JavaScript code base. Storing and sharing the different scenarios, with the ability to rollback changes.



Gaussian processes.Theory and applications in predictive modeling of spatiotemporal phenomena.

A basic introduction to Gaussian Processes and the mathematics behind them as well as some example applications.

Generating disparity maps using Convolutional Neural Networks

Extract 3D information from a single static 2D image, using CNNs and computer-generated dataset.

Intuitive Physics: Understanding Dynamics in Real World Physical Scenes

Using machine learning and physics engines to build models able to better understand dynamics in the surrounding world.

Recent & Upcoming Talks

Robot Learning Using Physics-Informed Models

Гаусови процеси - когато невронните мрежи не са достатъчни (Gaussian Processes - when neural networks are not enough)

С развитието на машиното самообучение и поредната вълна от интерес към невронните мрежи, интелигентни системи от голям мащаб като …

Gaussian Processes. Theory and Applications


  • Bayes Centre, 47 Potterrow, Edinburgh, Scotland, EH8 9BT
  • DM Me