Reinforced Vision-and-Language Navigation Based on Historical BERT

Vision-and-language navigation(VLN) task needs an agent navigate to a target location under the guidance of natural language instructions given by humans in novel 3D environments. We mainly study the vision-and-language navigation task in indoor scenes. It is difficult to obtain a large number of labeled training data, thus traditional multi-modal methods perform poorly in the data-scarce environment. We introduce pre-trained BERT model into vision-and-language navigation task to extract environment features and reduce the computation consumption by fixing the instruction features during navigation. Meanwhile, we design Gated Recurrent Vision-and-Language BERT to effectively memorize and transmit historical information to alleviate memory loss problem during navigation. In addition, we conduct reinforcement learning and imitation learning simultaneously to address the exposure bias of imitation learning. Empirically, we conduct experiments in discrete environment and analyze experimental results to verify the effectiveness of our proposed method.