Scene Memory Transformer for Embodied Agents in Long-Horizon Tasks
arXiv:1903.03878
Abstract
Many robotic applications require the agent to perform long-horizon tasks in partially observable environments. In such applications, decision making at any step can depend on observations received far in the past. Hence, being able to properly memorize and utilize the long-term history is crucial. In this work, we propose a novel memory-based policy, named Scene Memory Transformer (SMT). The proposed policy embeds and adds each observation to a memory and uses the attention mechanism to exploit spatio-temporal dependencies. This model is generic and can be efficiently trained with reinforcement learning over long episodes. On a range of visual navigation tasks, SMT demonstrates superior performance to existing reactive and memory-based policies by a margin.
arXiv commentCVPR 2019 paper with supplementary material
References in corpus (10)
- On the difficulty of training Recurrent Neural Networks
- On Evaluation of Embodied Navigation Agents
- Learning to Navigate in Complex Environments
- Control of Memory, Active Perception, and Action in Minecraft
- Predicting Depth, Surface Normals and Semantic Labels with a Common Multi-Scale Convolutional Architecture
- Neural SLAM: Learning to Explore with External Memory
- Generating Wikipedia by Summarizing Long Sequences
- Unifying Map and Landmark Based Representations for Visual Navigation
- Memory Augmented Control Networks
- Active Neural Localization