MSc Project: Active Inference for Robot Control
Cognitive Robotics Department, TU Delft
Active inference, also referred to as the free energy principle, is a framework formulated by K. Friston  to explain the behaviour of biological system as a minimisation of a free energy functional of their internal states exploiting beliefs about hidden states in environment. Active inference is explored in neuroscience for a unified account of perception and action, and has been proposed as an alternative to reinforcement learning or classical control. Approaches to generate optimised behaviour policies . Preliminary results have been obtained applying this framework in robot control .
The project will explore the possibilities of the biologically inspired active inference framework for the control of industrial manipulators. Dynamic control of industrial manipulators present challenges such as solving the inverse problem or sensory motor delays. This project will:
- Analyse the applicability of active inference to address challenges in dynamic robot control.
- Develop robot control solutions based on active inference.
- Create a ROS package for the implementation of active inference-based controllers for industrial manipulators.
The project will build on the results of  to first understand the potential value of the active inference framework for the control of robots, and then explore its applications:
- Designing experimental robotic setups in simulation and with a real robot arm to test the framework,
- Design predictive models under the active inference paradigm for relevant robot control applications, such as visual servoing, and
- Compare its performance to existing robot control approaches, evaluating amongst others:
- Robustness to exogenous perturbations (e.g. forces applied to the robot during movement)
- Robustness to deviations in the robot model
- Robustness to sensorimotor delays
 Friston, K., Kilner, J., & Harrison, L. (2006). A free energy principle for the brain. J Physiol Paris., 100 (1-3), 70-87.
 K. J. Friston, J. Daunizeau, and S.J. Kiebel. Reinforcement learning or active inference? PLOS ONE, 4(7):1-13, 07 2009.
 L. Pio-Lopez, A. Nizard, K. Friston, and G. Pezzulo. Active inference and robot control: a case study. J R Soc Interface, 13(122), Sep 2016.
To apply for this job email your details to email@example.com