OptoRoboRat III

Creation of a simulation environment for virtual behavioral experiments with rats, learning and explanation of animal behaviour, and realisation of a robotic rat


Relevant for Research Area

C - Applications

The project builds on

DeepDecode

PIs

Prof. Thomas Brox

Prof. Wolfram Burgard

Jun.-Prof. Joschka Boedecker


Summary

The aim of the project is the creation of a simulation environment for behavioural experimentswith a simulated rat, the development of a robot rat as a physical realization simulation, as well as the application and further development of methods of mechanical learning to model and analyse behavioural experiments. The simulation environment includes the simulation of the experimental environment and the rat's sensors/ senses from their egocentric perspective. This should be in harmony with the environment of the real animal so that similarities in behaviour between simulation and reality can be scientific conclusions regarding the perception and decision making of the realanimal can be drawn. Tracking data of real rats can be used as data basis for the simulation. In order to facilitate decision making by means of reinforcementlearning, the simulation should enable efficient learning experiments that are clearlycan be executed faster than real time. Through inverse reinforcement learning,and image-based navigation learning can be used to conduct experiments with real animalsbe reproduced and efficiently analyzed. Finally, the robot rat should be completelyenable observable experiments in the real world. Among other things, questions in reference to the transfer learning of behaviour learned in the simulation to the physicalrobots are examined.