SpaceDog is an initiative from The Space, an innovation center in Brussels, Belgium that aims to bridge emerging technologies and European policies for the citizens of Europe.
In that context, more specifically on the topic of Digital Learning, we wanted to promote innovation and creative thinking by challenging students to build a robot, initially based on the OpenDog by James Bruton, itself loosely based on the Boston Dynamics Spot Mini.
The challenges will be to make a version that can walk properly via remote control, within a 4-5 months time constrain, and within the 10K budget allocated to the project. It will also need to be designed in a way that can allow future upgrade, starting with getting rid of the remote, and letting walk on its own... That will be its own project, later down the line.
This project is open source is will be published a GNU GPLv3 license.
The hunt for a fitting linear actuator, that was both reasonably priced and shipped fast took a little while, but the team settled on a IP66 PA-04 model from Progressive Automations, the 6" stroke, 100lbs, 2.80"/s to be precise, which fit Laurent and Sam's initial calculations. Two of them were ordered.
In the mean time, they build the test bench out of 40x40 aluminium extrusion, with a ball bearing carriage to attach the leg to.
They also remade the CAD for the leg with the corresponding linear actuators, and designed all the joints and various other parts to be printed on the TAZ6 in nylon.
The leg itself makes use of two aluminium extrusion beams, a 40x20 for the lower part and a beefier 60x20 for the upper leg. The knee joint is all nylon for the prototype, but like most 3D printed part it will be upgraded later to machined aluminium most likely, with either a copper axel or ball bearings. The leg is mounted on a 50mm diameter tube, itself attached to the carriage on the test bench.
It will be controlled using an Arduino Mega, and powered by a 5000mAh LiPo battery, and using an AMT203 Encoder to measure rotation.
The team manage to assemble a first version of the leg, but the actuators turned out to be longer than what the spec sheet on the website advertised, which means the 300mm aluminium extrusion beams were too short.
New ones, 500mm ones were ordered instead, to be cut if necessary once they can validate the assembly.
After hours of printing and waiting for stuff to arrive, this is the result:
Currently, the guys are working on finalising the hardware on it. Due to being a bit behind schedule, the team settled on a PVC 50mm tube to carry the leg, which is proving to weak and slippery. Upgrades are being made, while Daniel started on the wiring, designing a new foot with more grip, using a flexible rubber like filament, while Laurent is running software test on the Arduino board.
The plan was to first analyse and retro-engineer the template for this robot, the OpenDog, in order to build an improved version, while still reusing as much as the open source assets available.
Thus the team did indeed start by watching hours of footage from OpenDog, going through the available files and brainstorming intensively on the overall design, and the components needed.
There were a few obstacles: James’ robot being a work in progress, the full part list is not yet available, and the files hosted in his GitHub repository were a start, but all the CAD files are only available in .stp format, and lack all joint constrains. Daniel started then by importing all the files into Inventor and adding back all the constrains.
The team then looked at brushless motors, starting with James' original ones, but finally decided on using linear actuators for several reasons: it was faster than building their own home made ones with brushless motors, it would help simplify the design and finally it is a challenge they liked, as there aren't many quadruped robots powered by linear actuators. Simplifying the design is a core principle of their approach, not only for the legs, but for the whole robot.
However they will use a yet to be chosen brushless motor for the hip and shoulder lateral rotations, connected to an axel on which the leg is mounted. Another departure from the OpenDog.
This all meant that the leg design had to be update to fit actuators. It was decided to keep the aluminium extrusion structure, but they redesign their own knee join, and attachement to the body. They also elected to keep an architecture based around Arduinos for now, each leg having an Arduino of its own, being controlled by a main Arduino board. The main processor will however need to be upgraded for something more powerful once the robot proves stable enough to walk.
In order to test their idea, they are building a test bench and a first leg prototype.