GoalVLM: VLM-driven Object Goal Navigation for Multi-Agent System

TL;DR

GoalVLM introduces a multi-agent framework that leverages vision-language models and spatial reasoning for zero-shot, open-vocabulary object navigation, outperforming prior methods without task-specific training.

Contribution

It integrates VLMs with spatial and semantic reasoning for multi-agent object navigation, enabling zero-shot generalization to novel goals.

Findings

Achieves 55.8% subtask success rate on GOAT-Bench

Outperforms state-of-the-art methods without training

Validates importance of VLM-guided reasoning and localization

Abstract

Object-goal navigation has traditionally been limited to ground robots with closed-set object vocabularies. Existing multi-agent approaches depend on precomputed probabilistic graphs tied to fixed category sets, precluding generalization to novel goals at test time. We present GoalVLM, a cooperative multi-agent framework for zero-shot, open-vocabulary object navigation. GoalVLM integrates a Vision-Language Model (VLM) directly into the decision loop, SAM3 for text-prompted detection and segmentation, and SpaceOM for spatial reasoning, enabling agents to interpret free-form language goals and score frontiers via zero-shot semantic priors without retraining. Each agent builds a BEV semantic map from depth-projected voxel splatting, while a Goal Projector back-projects detections through calibrated depth into the map for reliable goal localization. A constraint-guided reasoning layer evaluates frontiers through a structured prompt chain (scene captioning, room-type classification, perception gating, multi-frontier ranking), injecting commonsense priors into exploration. We evaluate GoalVLM on GOAT-Bench val_unseen (360 multi-subtask episodes, 1032 sequential object-goal subtasks, HM3D scenes), where each episode requires navigating to a chain of 5-7 open-vocabulary targets. GoalVLM with N=2 agents achieves 55.8% subtask SR and 18.3% SPL, competitive with state-of-the-art methods while requiring no task-specific training. Ablation studies confirm the contributions of VLM-guided frontier reasoning and depth-projected goal localization.