Stretch: A Small, Capable, and Affordable Mobile Manipulator

Stretch: A Small, Capable, and Affordable Mobile Manipulator

Charlie C. Kemp, Bilsen D. Kim, Blake S. Kim, Aaron Edsinger, Thomas Tao, Andreea Danielescu, Todor Davchev, Aran Sena, Tirtha Mandal, Adam Allevato · Hello Robot, Inc. · 2023

Framework

ROS 2 / Python

License

Apache-2.0

Stars

1,200

Summary

A low-cost mobile manipulator designed for indoor home and office environments, with a large open-source ecosystem and teleoperation-based data collection.

Abstract Summary

Stretch is a compact, mobile manipulator developed by Hello Robot with the explicit goal of making home and office robotics research affordable and scalable. At roughly the size of a small floor lamp, Stretch combines a lightweight telescoping arm on a differential-drive mobile base with a gripper and a mast-mounted Intel RealSense camera. Its retail price of approximately $25,000 makes it one of the most accessible full mobile-manipulation platforms for academic research, particularly for tasks in human-centric indoor environments. The hardware design emphasizes safety and simplicity for domestic settings. The arm is deliberately low-mass and low-power, making it safe to operate near people without heavy guarding. The mobile base is narrow enough to pass through standard doorways, and the telescoping arm provides a large vertical workspace from floor to countertop height. Onboard compute is provided by a Jetson Orin module, enabling local visual perception and neural-network inference. The robot runs ROS 2 natively, with all drivers, teleoperation interfaces, and calibration routines available as open-source packages. One of the primary use cases for Stretch is data-driven manipulation in real homes. Researchers use Stretch to collect large-scale datasets of daily household tasks—opening drawers, picking up objects, wiping surfaces—using a combination of teleoperation and kinesthetic demonstration. The resulting datasets feed into imitation learning and VLA pipelines. Several recent works (including contributions from the DROID dataset and open-source community) feature Stretch as one of the primary embodiment platforms because of its reproducibility and low barrier to deployment. The software ecosystem is another major strength. Hello Robot maintains well-documented Python APIs for low-level joint control, trajectory generation, and camera streaming. Integration with LeRobot, ACT, and diffusion policy frameworks is straightforward, and the community actively shares trained checkpoints and task-specific dataset collection scripts. For robotics labs aiming to study long-horizon mobile manipulation in unstructured environments, Stretch offers a pragmatic combination of hardware, cost, and software support that is difficult to match with custom-built platforms.

Key Points

  • Compact, safe mobile manipulator designed for indoor home/office tasks.
  • Low-cost platform (~$25k) with ROS 2, Python SDK, and teleoperation data collection.
  • Telescoping arm spans floor-to-countertop workspace; narrow base navigates homes.
  • Widely used in home robotic manipulation datasets and imitation learning research.
  • Large open-source community with integrations for LeRobot, ACT, and VLA stacks.

Additional Notes

Setup Tips

  • ROS 2 Humble is recommended for the latest Stretch driver support.
  • Use the provided calibration scripts to align depth and RGB streams before collecting datasets.
  • The arm is intentionally back-drivable; gentle guiding is possible for kinesthetic demos.
  • DROID (Khazatsky et al., 2024)
  • LeRobot (Cadene et al., 2024)
  • Mobile ALOHA (Fu et al., 2024)
Share

Related Papers

AIR-VLA+: Decoupling Movement and Manipulation via Cascaded Dual-Action Decoders with Asymmetric MoE for Aerial Robots

AIR-VLA+: Decoupling Movement and Manipulation via Cascaded Dual-Action Decoders with Asymmetric MoE for Aerial Robots

Jianli Sun, Bin Tian, Qiyao Zhang et al. · arXiv preprint · Jun 2026

Aerial manipulation systems have long suffered from representation coupling in end-to-end control, as platform-level Unmanned Aerial Vehicle (UAV) movement and end-effector-level arm manipulation differ substantially in action scale, dynamics, and control objectives. In this paper, we propose AIR-VL...

uav vla manipulation
PDF Advanced
No code repo Jun 2026
EgoForce: Forearm-Guided Camera-Space 3D Hand Pose from a Monocular Egocentric Camera

EgoForce: Forearm-Guided Camera-Space 3D Hand Pose from a Monocular Egocentric Camera

Christen Millerdurai, Shaoxiang Wang, Yaxu Xie et al. · SIGGRAPH 2026 · May 2026

Reconstructing the absolute 3D pose and shape of the hands from the user's viewpoint using a single head-mounted camera is crucial for practical egocentric interaction in AR/VR, telepresence, and hand-centric manipulation tasks, where sensing must...

manipulation manipulation simulation
Code PDF Intermediate
Code ★ 0 May 2026
MonoDuo: Using One Robot Arm to Learn Bimanual Policies

MonoDuo: Using One Robot Arm to Learn Bimanual Policies

Sandeep Bajamahal, Lawrence Yunliang Chen, Toru Lin et al. · ICRA 2026 · May 2026

Bimanual coordination is essential for many real-world manipulation tasks, yet learning bimanual robot policies is limited by the scarcity of bimanual robots and datasets. Single-arm robots, however,

manipulation bimanual imitation-learning dataset
PDF Intermediate
No code repo Code updated: May 2026
SI-Diff: A Framework for Learning Search and High-Precision Insertion with a Force-Domain Diffusion Policy

SI-Diff: A Framework for Learning Search and High-Precision Insertion with a Force-Domain Diffusion Policy

Yibo Liu, Stanko Oparnica, Simon Shewchun-Jakaitis et al. · arXiv · May 2026

Contact-rich manipulation is fundamental in robotics but poses significant challenges due to uncertainties in relative poses, such as misalignments and small clearances in peg-in-hole tasks. Existing approaches typically address search and...

manipulation tactile manipulation
Code PDF Intermediate
Code ★ 0 May 2026