MVSSC: Meta-reinforcement learning based visual indoor navigation using multi-view semantic spatial context

Wanruo Zhang, Hong Liu, Jianbing Wu, Yidi Li

Abstract: Highlights•Utilize the multi-view representation to enlarge the agent’s perception.•Construct a semantic graph and spatial grids to learn multi-view context information.•Incorporate meta-learning into the multi-view framework for better generalization.•Results show the superior generalization to new objects and new scenes.