Deep Active Localization

Abstract

Active localization is the problem of generating robot actions that allow it to maximally disambiguate its pose within a reference map. Traditional approaches to this use an information-theoretic criterion for action selection and hand-crafted perceptual models. In this work we propose an end-to-end differentiable method for learning to take informative actions that is trainable entirely in simulation and then transferable to real robot hardware with zero refinement. The system is composed of two modules: a convolutional neural network for perception, and a deep reinforcement learned planning module. We introduce a multi-scale approach to the learned perceptual model since the accuracy needed to perform action selection with reinforcement learning is much less than the accuracy needed for robot control. We demonstrate that the resulting system outperforms using the traditional approach for either perception or planning. We also demonstrate our approaches robustness to different map configurations and other nuisance parameters through the use of domain randomization in training. The code is also compatible with the OpenAI gym framework, as well as the Gazebo simulator.

arXiv comment10 pages

References in corpus (5)

• Deep Residual Learning for Image Recognition
• Densely Connected Convolutional Networks
• Markov Localization for Mobile Robots in Dynamic Environments
• Scalable trust-region method for deep reinforcement learning using Kronecker-factored approximation
• Domain Randomization for Transferring Deep Neural Networks from Simulation to the Real World