GridMM: Grid Memory Map for Vision-and-Language Navigation
This work addresses the challenge of environment representation for agents in vision-and-language navigation, offering a novel memory structure that enhances navigation accuracy in both discrete and continuous 3D environments.
The paper tackles the problem of representing previously visited environments in vision-and-language navigation by introducing GridMM, a top-down egocentric and dynamically growing grid memory map, which improves spatial relation representation and incorporates fine-grained visual clues. The method demonstrates superiority across multiple datasets including REVERIE, R2R, SOON, and R2R-CE, with concrete performance gains reported in experiments.
Vision-and-language navigation (VLN) enables the agent to navigate to a remote location following the natural language instruction in 3D environments. To represent the previously visited environment, most approaches for VLN implement memory using recurrent states, topological maps, or top-down semantic maps. In contrast to these approaches, we build the top-down egocentric and dynamically growing Grid Memory Map (i.e., GridMM) to structure the visited environment. From a global perspective, historical observations are projected into a unified grid map in a top-down view, which can better represent the spatial relations of the environment. From a local perspective, we further propose an instruction relevance aggregation method to capture fine-grained visual clues in each grid region. Extensive experiments are conducted on both the REVERIE, R2R, SOON datasets in the discrete environments, and the R2R-CE dataset in the continuous environments, showing the superiority of our proposed method.