Stanford Exoskeleton Boot

Not soldA university research prototype with no price. The CAD files and bill of materials are published with the Nature paper, and Stanford said in 2022 it hoped to see commercial versions in the coming few years.

Not for sale (research)SDK or research modeAnkle exoskeleton1.2 kg per ankle17% less energyOpen CAD (Nature)

Not for saleNot sold. It is a university research prototype; labs can build their own from the published design files. Nature paper ↗

Control: The Raspberry Pi controller is programmed by the researchers, and the published code and designs let other labs build and run their own copy.

Stanford Engineering news ↗Nature paper (open access) ↗Stanford Biomechatronics Lab ↗

Overview

The Stanford exoskeleton boot is a research prototype from Steve Collins' Biomechatronics Lab, an untethered ankle exoskeleton that personalizes its own assistance while the wearer walks in everyday settings. Published in Nature on 2022-10-12 (Slade et al.), it used a machine-learning model trained on years of tethered emulator experiments to tune each person's assist within about an hour, giving 9% faster walking and 17% less energy per distance in 10 participants. The hardware is simple for its class, a CubeMars AK80-9 motor, a Raspberry Pi 4B and a small battery in a 1.2 kg unit per ankle, and the design files are published under CC BY 4.0. It was never sold, and no spin-off product based on it has been announced (unconfirmed). Commercial relatives include the Dephy ExoBoot and consumer Sidekick, while its control research came from Caplex-style emulators.

Specifications

Category	Exoskeletons & exosuits · Full-body, military & research
Released	2022
Country	USA
Availability	Not for sale (research)
Programmability	Open source (hardware + software)
Built-in autonomy	Reactive · Responds to its surroundings without a map, such as avoiding obstacles or following a person
LLM support	None · No voice assistant or LLM.
Vision sensors	No vision sensors
Actuators	Brushless / direct-drive motors
Audio	No audio
Compute type	Other ARM chip
Control	SDK or research mode
Weight	1.2 kg per ankle (battery 0.3 kg)
Assistance	A brushless motor at the heel pulls on the boot during push-off with up to 54 Nm at 1.5 m/s. In outdoor tests it raised walking speed by 9% and cut energy per distance by 17%, which Stanford likened to taking off a 30 lb (14 kg) backpack.
Joints	Ankles
Battery & runtime	24 V, 1,300 mAh lithium-polymer battery (0.3 kg), at least 30 minutes of walking per charge.
Operation	No wearer controls. A machine-learning model trained on years of emulator data adjusts the assistance pattern to each person within about an hour of normal walking.
Speed	Tested at self-selected walking speeds and 1.5 m/s on a treadmill
Sensors	Ankle rotary encoder, pressure-sensing insole, strain-gauge torque sensor on the heel spur.
Compute	Raspberry Pi 4B running control and optimization at 200 Hz
SDK	CAD files, bill of materials and Python optimization code published as Nature supplementary data