[Paper] YOR: Your Own Mobile Manipulator for Generalizable Robotics

Source: arXiv - 2602.11150v1

Overview

The paper presents YOR (Your Own Robot), an open‑source, low‑cost mobile manipulator that packs an omnidirectional base, a telescopic lift, and two articulated arms with grippers into a single, modular platform. By keeping the bill‑of‑materials under $10 k, the authors show that sophisticated whole‑body mobility and bimanual manipulation are no longer confined to expensive research labs, opening the door for a broader community of developers to experiment with advanced robot learning and autonomy.

Key Contributions

• Fully open‑source hardware design (CAD files, wiring diagrams, and parts list) enabling anyone to build the robot from off‑the‑shelf components.
• Modular architecture that separates the base, lift, and dual‑arm subsystems, simplifying upgrades and repairs.
• Omnidirectional base + telescopic lift that together provide 6‑DoF whole‑body mobility, a rare combination at this price point.
• Dual‑arm bimanual capability with interchangeable grippers, supporting coordinated manipulation tasks.
• Demonstrated benchmark tasks (autonomous navigation, whole‑body pose control, and bimanual object handling) that match or exceed the performance of many commercial platforms.
• Comprehensive software stack built on ROS 2, including perception, planning, and control modules ready for plug‑and‑play use.

Methodology

The authors approached the design problem in three stages:

1. Component Selection & System Integration – They surveyed commercially available actuators, sensors, and computing boards, choosing parts that balance cost, reliability, and ease of sourcing. The omnidirectional base uses three mecanum wheels driven by brushless DC motors, while the lift employs a linear actuator with a built‑in encoder for precise height control.

2. Modular Mechanical Design – Using parametric CAD models, each subsystem (base, lift, arms) can be printed or CNC‑machined independently. Fasteners and standard connectors allow quick assembly without specialized tools.

3. Software Architecture – A ROS 2‑based framework ties together perception (RGB‑D cameras, LiDAR), planning (MoveIt 2 for arm trajectories, Nav2 for navigation), and whole‑body control (a custom inverse‑kinematics solver that treats the lift and base as additional joints). The stack is containerized for reproducible deployment on a single onboard PC (e.g., an NVIDIA Jetson AGX).

The team validated the platform by programming a set of representative tasks that stress different capabilities: navigating to a target location, raising the lift to reach a shelf, and using both arms to pick up and hand over an object.

Results & Findings

*Numbers are taken from publicly available specs of platforms such as the Fetch robot and the Toyota HSR.

Key takeaways

• Cost‑performance ratio: YOR delivers comparable navigation and manipulation success rates at a fraction of the price.
• Whole‑body coordination: The integrated lift and omnidirectional base enable smooth, collision‑free motions that would otherwise require separate planning pipelines.
• Ease of replication: All hardware files and build instructions allowed the authors’ second lab to reproduce the robot in under two weeks, confirming the design’s reproducibility.

Practical Implications

• Rapid prototyping for robot‑learning research – Developers can now iterate on perception‑action pipelines (e.g., vision‑based grasping, reinforcement learning) without waiting for expensive hardware.
• Education & training – Universities and bootcamps can afford to equip labs with multiple YOR units, giving students hands‑on experience with mobile manipulation.
• Industry pilots – Small‑to‑medium enterprises can test warehouse pick‑and‑place, inventory scanning, or service‑robot use cases before committing to high‑end platforms.
• Open‑source ecosystem growth – The ROS 2‑centric software stack encourages community contributions (new grippers, sensor packages, or higher‑level skill libraries), accelerating innovation.
• Modular upgrades – Because each subsystem is independent, teams can swap in higher‑torque arms, better cameras, or more powerful compute boards as budgets allow, extending the robot’s lifespan.

Limitations & Future Work

• Payload constraints – The current arms are limited to ~1 kg, restricting manipulation of heavier objects; future revisions could integrate higher‑capacity actuators.
• Robustness in harsh environments – The design targets indoor labs; dust‑proofing, waterproofing, and more rugged chassis are needed for field deployments.
• Scalability of control – Whole‑body inverse‑kinematics become computationally heavier as more degrees of freedom are added; the authors plan to explore learning‑based controllers to maintain real‑time performance.
• User studies – While benchmark tasks were demonstrated, systematic usability studies with developers of varying expertise are pending to quantify the learning curve.

The authors envision YOR as a living platform that will evolve through community feedback, gradually closing the gap between low‑cost hobbyist robots and high‑end research manipulators.

Authors

• Manan H Anjaria
• Mehmet Enes Erciyes
• Vedant Ghatnekar
• Neha Navarkar
• Haritheja Etukuru
• Xiaole Jiang
• Kanad Patel
• Dhawal Kabra
• Nicholas Wojno
• Radhika Ajay Prayage
• Soumith Chintala
• Lerrel Pinto
• Nur Muhammad Mahi Shafiullah
• Zichen Jeff Cui

Paper Information

• arXiv ID: 2602.11150v1
• Categories: cs.RO, cs.LG
• Published: February 11, 2026
• PDF: Download PDF