Research Analyzer
← Back ICRA 2026

Can a Robot Walk the Robotic Dog: Triple-Zero Collaborative Navigation for Heterogeneous Multi-Agent Systems

Yaxuan Wang, Yifan Xiang, Ke Li, Xun Zhang, BoWen Ye, Zhuochen Fan, Fei Wei, Tong Yang

PDF

AI summary

Key figure (auto-extracted from paper)
TZPP enables a humanoid and quadruped robot to collaboratively navigate complex real-world environments without training, prior maps, or simulation, achieving human-comparable efficiency.
Triple-Zero Navigation Heterogeneous Multi-Robot Systems LLM-Guided Navigation Real-World Deployment Cooperative Path Planning Zero-Shot Generalization

Problem

Existing multi-robot path planning methods heavily rely on large-scale training, prior environmental maps, or simulated environments, which limits their generalization and practical deployment in dynamic, unknown real-world settings.

Approach

A coordinator-explorer framework where a humanoid robot handles high-level task coordination while a quadruped robot explores and identifies feasible paths, all guided by a multimodal large language model without any prior training or simulation.

Key results

  • First heterogeneous path planning paradigm satisfying zero training, zero prior knowledge, and zero simulation constraints
  • Human-level collaborative efficiency and navigation accuracy across diverse indoor and outdoor real-world scenarios
  • Adaptive dual-mode exploration strategy that robustly handles both landmark-sparse and obstacle-rich environments
  • Validated real-world deployment on Unitree G1 and Go2 robots demonstrating strong zero-shot generalization

Why it matters

Eliminates simulation-to-reality gaps and high training costs, accelerating the practical deployment of intelligent heterogeneous robot teams in dynamic real-world applications.

Abstract

We present Triple Zero Path Planning (TZPP), a collaborative framework for heterogeneous multi-robot systems that requires zero training, zero prior knowledge, and zero simulation. TZPP employs a coordinator–explorer architecture: a humanoid robot handles task coordination, while a quadruped robot explores and identifies feasible paths using guidance from a multimodal large language model. We implement TZPP on Unitree G1 and Go2 robots and evaluate it across diverse indoor and outdoor environments, including obstacle- rich and landmark-sparse settings. Experiments show that TZPP achieves robust, human-comparable efficiency and strong adaptability to unseen scenarios. By eliminating reliance on training and simulation, TZPP offers a practical path toward real-world deployment of heterogeneous robot cooperation. Our code and video are provided at: https://github.com/triple- zeropp/Triple-zero-robot-agent.

Index terms

Cooperating Robots Multi-Robot Systems Path Planning for Multiple Mobile Robots or Agents

Related papers