DoShiCo: a Domain Shift Challenge for Control.

Training deep neural control networks end-to-end for real-world applications typically requires big demonstration datasets in the real world or big sets consisting of a large variety of realistic and closely related 3D CAD models. These real or virtual data should, moreover, have very similar characteristics to the conditions expected at test time. These stringent requirements and the time consuming data collection processes that they entail, are probably the most important impediment that keeps deep neural policies from being deployed in real-world applications. Therefore, in this work we advocate an alternative approach, where instead of avoiding any domain shift by carefully selecting the training data, the goal is to learn a policy that can cope with it. To this end, we propose a new challenge: to train a model in very basic synthetic environments, far from realistic, in a way that it can fly in more realistic environments as well as take the control decisions on real-world data. We collected a benchmark dataset and implemented a baseline method, exploiting depth prediction as an auxiliary task to help overcome the domain shift. Even though the policy is trained in very basic environments, it can learn to fly in a very different realistic simulated environment. It is even capable to compete and in some cases outperform a policy trained in the more realistic environment when testing on real-world data.