Pernilla Craig

Email: pernilla.craig@brl.ac.uk

Supervisors: Nathan Lepora

PhD Project Title: Active Inference for Tactile Sensor Control

Research Group: Tactile Robotics

Biography

I moved into the field of robotics after graduating with a Masters in Physics from the University of Exeter, and am currently in the second year of my PhD working as part of the tactile sensing research group. Tactile sensing is the ability to not only touch but to feel the world, allowing us to grasp objects with precision and identify objects without having to look at them. This ability is able to offer significant advantages to robotics and has the potential to greatly improve the interactions robots are able to have with the world.

I’m currently working on creating an intelligent control system for tactile sensors using a neuroscience theory regarding the human brain and how it processes action and perception whilst making decisions. This is believed to be through the mechanism of prediction error minimisation. As such this work involves creating a predictive belief system for the ‘sensing agent’ that can compare the expected outcome of an action with the actual outcome that is observed after it has been completed. By enacting this process and learning about its surroundings the agent can improve its internal model of the world which in turn allows for better interactions in the future. The sensing agent keeps a history of its beliefs and updates them based upon new evidence, either solidifying them if it agrees or changing them if it contradicts. The aim here is for the completed system to be able to learn about its surroundings and choose optimal sequences of actions to achieve its overall goal. This goal can be any one of a large number of objectives, such as locating itself within a system, following a contour, or identifying an object.