PAINT: Partner-Agnostic Intent-Aware Cooperative Transport with Legged Robots

Zhihao Cao; Tianxu An; Chenhao Li; Stelian Coros; Marco Hutter

cs.ROApr 14, 2026

PAINT: Partner-Agnostic Intent-Aware Cooperative Transport with Legged Robots

Authors: Zhihao Cao, Tianxu An, Chenhao Li, Stelian Coros, Marco Hutter

Organizations: Robotic Systems Lab; ETH Zurich, Switzerland · Computational Robotics Lab; ETH Zurich, Switzerland · ETH AI Center; Switzerland

Abstract

Collaborative transport requires robots to infer partner intent through physical interaction while maintaining stable loco-manipulation. This becomes particularly challenging in complex environments, where interaction signals are difficult to capture and model. We present PAINT, a hierarchical learning framework for partner-agnostic intent-aware collaborative legged transport that represents partner intent as an explicit interaction wrench and recovers it from payload-coupled proprioceptive histories. PAINT decouples intent understanding from terrain-robust locomotion: A high-level policy uses the inferred interaction wrench for transport and reconstructs it through teacher-student training, while a low-level locomotion backbone ensures robust execution. This enables lightweight deployment without external force-torque sensing or payload tracking. Extensive simulation and real-world experiments demonstrate compliant cooperative transport across diverse terrains, payloads, and partners. Furthermore, we show that PAINT can reuse the same single-agent policy for decentralized multi-robot transport by mechanically combining team interactions into a common wrench space, and supports heterogeneous-team transport. Our results suggest that payload-coupled proprioceptive interaction provides a physically grounded interface for partner-agnostic intent-aware collaborative transport.

Figures & tables

Fig. 1 : Partner-agnostic intent-aware cooperative transport via purely proprioceptive feedback. Our framework enables legged robots to transport a shared payload with diverse partners across various terrains. Relying solely on proprioceptive signals, the robot infers the partner’s intent without external force-torque sensors or explicit payload pose tracking during transport. This lightweight approach ensures robust, compliant coordination across heterogeneous transport team configurations and can scale to decentralized multi-robot transport. The project website is available at https://paint-bot.github.io .

Fig. 2 : Overview of the partner-agnostic intent-aware cooperative transport framework. The system contains privileged training in simulation and lightweight real-world deployment via a hierarchical controller: (A) Intent Generator: Random forces and torques are applied to the payload to simulate partner guidance in simulation. (B) High-Level Controller: Trained with teacher-student scheme, a HL policy maps intent to base commands and arm joint actions, where a teacher (B1) uses wrench observations, and a deployable student (B2) relies on proprioceptive histories with an estimator to infer interaction wrench; KL regularization (B3) distills the teacher’s action distribution into the student. (C) Low-Level Controller: A terrain-robust locomotion backbone tracks the HL commands for stable transport under load. (D) Evaluation: The framework is developed in simulation (D1) and transferred to real-world deployment (D2) without FT sensing or payload tracking.

Reward Term	Expression	Weight ( $\times dt^{\#}$ )
Force Tracking	Eq. ( 11 )	$-2.0\times 10^{-2}$
Torque Tracking	Eq. ( 12 )	$-1.0\times 10^{-2}$
Payload Height	Eq. ( 13 )	$-3.0\times 10^{-3}$
Joint torque penalty	$\sum_{i}\tau_{i}^{2}$	$-2.0\times 10^{-9}$
Joint DoF penalty	$\sum_{i}\dot{q}_{i}^{2}+0.00025\,\ddot{q}_{i}^{2}$	$-2.0\times 10^{-4}$
Action penalty	$\sum_{i}(a_{t,i}-a_{t-1,i})^{2}+0.5\,a_{t,i}^{2}$	$-2.0\times 10^{-4}$

TABLE I : Reward For Learning the Transport Policy

Panel of Figure 4 — Fig. 3 : Overview of partner-agnostic intent-aware cooperative transport results. The proposed intent-aware framework enables stable cooperative transport across diverse terrains, payloads, partners, robot teams, and robot embodiments.

Fig. 4 : Real-world transport statistics across diverse leaders, payloads, and terrain. (a) The shared payload stays near level and (b) the partner-intent alignment remains high, with (c) per-condition statistics (success, payload tilt, SPARC smoothness, and intent alignment) summarized in the table.

Fig. 5 : Simulation analysis of scalability. Partner-intent alignment is evaluated for teams of one to four robots under varying (a) payload masses ( $12$ - $28\,$ kg) and (b) payload geometries. A single robot degrades sharply with increasing payload mass, whereas three- and four-robot teams sustain high alignment, showing improved handling of heavy payload-coupled dynamics.

Fig. 6 : Real-world multi-robot coordination with one human and two robots carrying an $8$ kg long box over two repeated trials: a mostly straight (Trial 1) and a more turning-intensive (Trial 2). (a) Inter-robot spacing remains tightly distributed (CoV $\approx 0.1$ ), and (b) the shared payload remains near level, indicating consistent decentralized coordination through payload-coupled interaction.

Fig. 7 : Real-world cross-embodiment transport with an ANYmal. While carrying a long box, the robot follows the human leader and maintains (a) small payload tilt and (b) steady carrying speed and heading, with (c) a tightly concentrated tilt distribution and (d) a concentrated carrying-speed distribution, demonstrating stable transport from proprioception alone.

Fig. 8 : The HL policy reacts purely to proprioceptive signals without active obstacle avoidance. When approaching obstacles (left), it cannot proactively adjust its motion, leading to collisions (middle) and causing locomotion failure (right).

Fig. 9 : Saliency analysis of the intent estimator under a sequential interaction schedule across arm joint positions.

Multi-Agent Collaboration

Multi-agent collaboration studies how autonomous agents coordinate, share information, and combine capabilities to solve tasks together.

See all 312 papers

Learning from Humans for Proactive Assistance in Human-Robot Collaborative Transport
Sep 21, 2026Elvin Yang, Christoforos MavrogiannisRobotic Control Compliant Robot Control
Department of Robotics, University of Michigan, Ann Arbor, United States
HA-DiPO: Diffusion Policy Optimization for Heterogeneous Cooperative Transport
Oct 24, 2026Xuebin Li, Jinpeng Xiao, Zong Chen +4Multi-Robot Cargo Transportation Cooperative MARL
School of Mathematics, Harbin Institute of Technology, Harbin, China · State Key Laboratory of Intelligent Manufacturing Equipment and Technology, School of Mechanical Science and Engineering, Huazhong University of Science and Technology, Wuhan, China · Department of Electrical Engineering and Information Technology, University of Naples Federico II, Naples, Italy
CoBrush: A Hierarchical Planning Framework for Human-Robot Co-Painting
Sep 28, 2026Dantong Qin, Yike Guo, Qinlin Liu +2Human-Robot Collaboration Robot Task Planning
Faculty of Industrial Design Engineering, Delft University of Technology, 2628 CD Delft, The Netherlands · Department of Computer Science and Engineering, Electronic and Computer Engineering, Hong Kong University of Science and Technology, Kowloon Hong Kong

Sep 21, 2026J/K move · Enter open · S save

Learning from Humans for Proactive Assistance in Human-Robot Collaborative Transport

Elvin Yang, Christoforos Mavrogiannis

Department of Robotics, University of Michigan, Ann Arbor, United States

Robotic Control Compliant Robot Control Human-Robot Collaboration Human Behavior Prediction

Open paper AskPDF arXiv

We focus on human-robot collaborative transport, a challenging task of broad relevance spanning logistics, manufacturing, and the home, in which a user and a robot work together to relocate a large or heavy object. To act as an effective partner, the robot should reduce the user's effort by contributing to efficient relocation of the object while remaining physically responsive to them. Prior work often addresses these capabilities separately, producing robots that may move the object efficiently but resist user input, or accommodate the user but depend on continuous guidance. Our key insight is that obstacle-constrained collaborative transport requires integrating predictions of human collaborative behavior with compliant robot control. To this end, we introduce PROACT, a framework for human-robot collaborative transport that incorporates anticipation into compliant whole-body control through a learned model of human collaborative behavior. Trained on a large-scale, real-world dataset of dyadic human transport demonstrations, our transformer architecture distills collaborative behavior into predictions of future object motion. Across 108 real-world trials with a 9-DoF mobile manipulator, PROACT reduces mean interaction work by 59.2% and 20.4%, and mean completion time by 12.9% and 6.9%, relative to compliance-only and MPC baselines, respectively. Footage from our experiments can be found at https://youtu.be/qAGvQfVPjbk.

Abstract

Multi-Agent Collaboration

Explore similar work

Learning from Humans for Proactive Assistance in Human-Robot Collaborative Transport

HA-DiPO: Diffusion Policy Optimization for Heterogeneous Cooperative Transport

CoBrush: A Hierarchical Planning Framework for Human-Robot Co-Painting