DoShiCo Challenge: Domain Shift in Control Prediction

Klaas Kelchtermans, Tinne Tuytelaars

Training deep neural network policies end-to-end for real-world applications so far requires big demonstration datasets in the real world or big sets consisting of a large variety…