GAPLE: Generalizable Approaching Policy LEarning for Robotic Object Searching in Indoor Environment

Abstract

We study the problem of learning a generalizable action policy for an intelligent agent to actively approach an object of interest in an indoor environment solely from its visual inputs. While scene-driven or recognition-driven visual navigation has been widely studied, prior efforts suffer severely from the limited generalization capability. In this paper, we first argue the object searching task is environment dependent while the approaching ability is general. To learn a generalizable approaching policy, we present a novel solution dubbed as GAPLE which adopts two channels of visual features: depth and semantic segmentation, as the inputs to the policy learning module. The empirical studies conducted on the House3D dataset as well as on a physical platform in a real world scenario validate our hypothesis, and we further provide in-depth qualitative analysis.

References in corpus (8)

• Playing Atari with Deep Reinforcement Learning
• DeCAF: A Deep Convolutional Activation Feature for Generic Visual Recognition
• Indoor Semantic Segmentation using depth information
• Building Generalizable Agents with a Realistic and Rich 3D Environment
• Sample Efficient Actor-Critic with Experience Replay
• Driving Policy Transfer via Modularity and Abstraction
• Analyzing Modular CNN Architectures for Joint Depth Prediction and Semantic Segmentation
• Divide-and-Conquer Reinforcement Learning