An Open Torque-Controlled Modular Robot Architecture for Legged Locomotion Research

TL;DR An open-source, torque-controlled quadruped (Solo) built from eight identical 3D-printed actuator modules with brushless motors and a 9:1 belt transmission. The 2.2 kg robot cost about 4000 € in materials, reaches a dimensionless leg stiffness of 10.8 and jumps to a base height of 65 cm.
What they do
Legged robots with good force control are usually expensive, hard to machine and not available to buy. This paper presents an open-source alternative: one low-cost torque-controlled actuator module, a foot contact sensor, a 2-DOF leg made from two modules, and the 8-DOF quadruped Solo made from four legs. Mechanical drawings, electronics and control software are all released under the BSD-3-clause license.
How it works
- Actuator module. A hobby brushless motor drives the joint through a 9:1 dual-stage timing belt, with an optical encoder on the motor shaft, all inside a 3D-printed shell. The low gear ratio lets joint torque be controlled from motor current alone, with no torque sensor. Only the motor shaft and pulleys need machining.
- Foot contact sensor. An LED and a light sensor are separated by a spring-loaded aperture that shifts under external force. It weighs 10 g and senses contact over a 270 degree range.
- Electronics. A custom driver board runs field-oriented torque control for two motors at 10 kHz; an off-board real-time PC closes the control loop at 1 kHz over CAN, Ethernet or WiFi.
- Controller. Motions come from a kino-dynamic optimizer. A quadratic program distributes foot forces to track the planned center of mass, angular momentum and base orientation, combined with a low-impedance controller on each leg.
Results
- The module weighs 150 g and outputs up to 2.7 Nm at 12 A. Solo weighs 2.2 kg and was built for approximately 4000 € of material cost.
- Leg stiffness can be commanded from 20 to 360 N/m without damping. Measured stiffness matches the command below 150 N/m and reaches about 266 N/m at the top, a dimensionless stiffness of 10.8; running humans are between 10 and 22.
- The contact sensor triggers at 3 N and detects contact about 3 ms after the reference force sensor, against about 31 ms when using motor currents.
- Solo walks over a seesaw that the planner does not know about, balances on moving platforms and jumps to a base height of 65 cm.
Limitations
Solo is tethered for power and data, and its eight joints only move in the sagittal plane. Measured stiffness falls below the commanded value at high stiffness, which the authors attribute to friction, transmission flexibility and encoders that measure the motor instead of the joint. The hobby motors show torque ripple at very low speed. Plans are tracked without re-planning.
Figures




Abstract
We present a new open-source torque-controlled legged robot system, with a low-cost and low-complexity actuator module at its core. It consists of a high-torque brushless DC motor and a low-gear-ratio transmission suitable for impedance and force control. We also present a novel foot contact sensor suitable for legged locomotion with hard impacts. A 2.2 kg quadruped robot with a large range of motion is assembled from eight identical actuator modules and four lower legs with foot contact sensors. Leveraging standard plastic 3D printing and off-the-shelf parts results in a lightweight and inexpensive robot, allowing for rapid distribution and duplication within the research community. We systematically characterize the achieved impedance at the foot in both static and dynamic scenarios, and measure a maximum dimensionless leg stiffness of 10.8 without active damping, which is comparable to the leg stiffness of a running human. Finally, to demonstrate the capabilities of the quadruped, we present a novel controller which combines feedforward contact forces computed from a kino-dynamic optimizer with impedance control of the center of mass and base orientation. The controller can regulate complex motions while being robust to environmental uncertainty.
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.
