Navi2Gaze: Leveraging Foundation Models for Navigation and Target Gazing

TL;DR

This work develops a VLM-driven method called Navigation-to-Gaze (Navi2Gaze) for efficient navigation and object gazing based on task descriptions that significantly outperforms existing approaches by precisely determining the optimal orientation relative to target objects.

Abstract

Task-aware navigation continues to be a challenging area of research, especially in scenarios involving open vocabulary. Previous studies primarily focus on finding suitable locations for task completion, often overlooking the importance of the robot's pose. However, the robot's orientation is crucial for successfully completing tasks because of how objects are arranged (e.g., to open a refrigerator door). Humans intuitively navigate to objects with the right orientation using semantics and common sense. For instance, when opening a refrigerator, we naturally stand in front of it rather than to the side. Recent advances suggest that Vision-Language Models (VLMs) can provide robots with similar common sense. Therefore, we develop a VLM-driven method called Navigation-to-Gaze (Navi2Gaze) for efficient navigation and object gazing based on task descriptions. This method uses the VLM to score and select the best pose from numerous candidates automatically. In evaluations on multiple photorealistic simulation benchmarks, Navi2Gaze significantly outperforms existing approaches by precisely determining the optimal orientation relative to target objects, resulting in a 68.8% reduction in Distance to Goal (DTG). Real-world video demonstrations can be found on the supplementary website

Figures (9)

• Figure 1: Workflow of Navi2Gaze: (1) Identification of target scene; (2) Reconstruction of task-aware space; (3) Generation of location candidates; (4) Sequential decisions for gazing objects.
• Figure 2: Framework of Navi2Gaze: 1. Identification of Target Scene: The robot uses GPT-4V to find and navigate to the target object using image sequences. 2. Reconstruction of Task-aware Space: The robot moves its camera to map the scene around the target object. 3. Generation of Location Candidates: Self-organizing Maps (SoM) and GPT-4V identify the target and suggest possible regions for the robot to position itself. 4. Sequential Decisions for Gazing Objects: GPT-4V scores these regions and directs the robot to the best position for observing the target.
• Figure 3: Assistive process of GPT-4V in Navi2Gaze.
• Figure 4: Method for Generating Candidate Regions.
• Figure 5: Mobile manipulation robot.