TinyARM-X: A Series of Open-Source, Configurable, and Transparent Torque-Controlled Desktop-Scale Robotic Manipulators

TL;DR An open-source family of 3D-printed, torque-controlled desktop arms (3 to 7 DoF, under 350 USD) built from small Dynamixel servos plus actuator and rigid-body dynamics models. The 7-DoF arm backdrives with 1.4 N of breakaway force and, teleoperated in mixed reality under Cartesian impedance control, presses target loads onto a scale within 7% error.
What they do
Cheap arms such as the SO-101 use stiff, position-controlled servos with high gear ratios, which makes them hard to backdrive, hard to simulate and risky in collisions. TinyARM-X is a family of 3D-printed desktop manipulators, from 3 to 7 DoF, that are torque-controlled instead. The full bill of materials for one arm with up to 7 DoF and a gripper is 346.65 USD.
How it works
- Common actuation unit. Every joint is a Dynamixel XL330 servo wrapped in a standard mount held by two screws. Adding or removing a DoF means swapping one link and one daisy-chained cable; the software counts the servos and loads the matching URDF.
- Actuator model. A fitted root model maps current command to output torque. A Viscous-Coulomb friction model is fed forward, and an alpha-beta filter estimates joint velocity from encoder positions read at about 1000 Hz.
- Whole-body control. Pinocchio computes gravity, Coriolis and friction compensation, with null-space control on the 7-DoF arm. Modes: a transparent mode for backdriving, joint and Cartesian impedance control, and haptic rendering that drives a MuJoCo model through god-object rendering.
- MR teleoperation. A Meta Quest 3 app sets the target pose, solved by a QP-based IK.
Results
- TinyARM-7: 209.8 mm safe reach, 33.5 g payload at full extension, 0.77 m/s maximum speed.
- Coupled to a Flexiv Rizon 10s with a force-torque sensor: 1.4 N backdrive breakaway force and 0.7 N running force. A commanded 40 N/m Cartesian stiffness comes out with 17.5 N/m average error, but at least 15 N/m on every axis.
- Swapping one joint takes 259.2 s on average with hand tools.
- Pressing 100, 150 and 200 g onto a scale through MR teleoperation: Cartesian impedance errs by 6.77%, -0.60% and 2.31%; joint impedance by up to -45.24%.
Limitations
Shoulder and elbow torque saturate when the arm is extended, which limits payload, stiffness accuracy and haptic feedback strength. High haptic gains can destabilize the system. The torque model is open-loop and ignores how speed affects maximum torque. The teleoperation and reconfiguration tests were each run by an expert user.
Figures




Abstract
Visuomotor policies have demonstrated that imprecise low-cost robots can autonomously complete complex manipulation tasks. However, most inexpensive manipulators rely on compact and rigidly position-controlled servos for actuation, which complicate simulation, operation, and safety. This work introduces TinyARM-X, a framework for easily manufacturable, configurable, and torque-controllable low-cost (<350 USD) serial manipulators. By using a single hardware/software architecture, a <7 Degree of Freedom (DoF) TinyARM can be manually reconfigured to an operational >3 DoF robot within 10 min. Actuator modeling and dynamic compensation enable torque-controlled applications like mixed-reality impedance control teleoperation and simulator force feedback rendering. Experiments include mechanical benchmarking and real-world force-aware telemanipulation. The TinyARM-7 has a payload of 33.5 g, a max speed of 0.77 m/s, and a reach of 210 mm. Moreover, with dynamic compensation, the robot has a backdrive-breakaway force of 1.4 N and a backdrive-running force of 0.7 N. Additionally, Cartesian-Impedance Control (CIC) attains a translational stiffness of >15 N/m. Lastly, the TinyARM-X stack is open-sourced to disseminate accessible desktop-scale torque-controlled robotic manipulation to the community.
Summaries are written automatically from each paper's text. Figures and videos come from the paper and its project page; all credit goes to the authors.
