Anyone who’s ever tried to build a bipedal robot will quickly start pulling their own hair out. There are usually a lot of servos involved, and controlling them all in a cohesive way is frustrating to say the least. [Mark] had this problem while trying to get his robot to dance, and to solve it he built a control system for a simple bipedal robot that helps solve this problem.
[Mark]’s robot has six servo motors per leg, for a total of 12 degrees of freedom. Commands are sent to the robot with an RC radio, and the control board that he built, called the Smart Servo Controller, receives the signals and controls the servos appropriately. There are 14 outputs for servos, operating at 12 bits and 50 Hz each, as well as 8 input channels. The servo controller can be programmed on a computer with user-selectable curves for various behaviors for each of the servos on the project. This eliminates the need to write cumbersome programs for simple robot movements, and it looks like it does a pretty good job!
Full disclosure: [Mark] currently has this project up on Kickstarter, but it is a unique take on complex robot control that could help out in a lot of different ways. Since you don’t need to code anything, it could lower the entry barrier for this type of project, possibly opening it up to kids or school projects. Beyond that, even veterans of these types of projects could benefit by not having to do as much brute-force work to get their creations up and moving around!
This reminded me a lot of Dog robot rom Half-Life 2 series.
That said, I wish he’d used something like Adafruit PWM expander ( https://www.adafruit.com/products/815 ). Would make this easily expandable, I guess.
I loved Half-Life2.
I’m Mark, the creator of this project. Adding I2C functionality is a good idea, I’ll take into consideration for a firmware update. One of the key ideas of this is simplicity and of not having to write interfaces and talk to registers in different devices. I know many people can program that easily but others can be intimidated by the task. :)
I think I’ll throw some pull requests once I get to building robots like this myself =) Your software is really nice when it comes to testing platform’s possibilities and calibration, or so it seems. I’m just not the type that buys pre-built boards (read it as: broke student ;-) )
that is so cool!
What about using AI to learn complex things like walking, grabbing objects, etc? Walking seems like a great place to start actually applying AI to robots. Get them to learn to walk on their own.
AI is actually a huge interest of myn, Machine learning, Optimization techniques. All projects has to start somewhere though. Baby robot steps right.
I’m going to make some OS printables templates to cut this kind of servo brackets from metal sheets using a heavy duty scissor. Zinc sheets are very affordable in my country because they use it for roofs and stuff. Think 2D and fold…
I’m sure it’s been done, and I’m sure it’s more complicated than it sounds, but what stops you from finding center of balance to tilt the hips left and right as the opposite leg lifts, shifting the weight on to the leg not moving? I see so many of these bipedal robots that scoot the feet, this seems to bypass the need to balance on one leg. Otherwise it would fall over. Maybe a simple acellerometer at the hips to determine center orientation?
Yes you are correct. I initially intended for this to completely lift the leg up but I found that the robot didn’t have enough mass to lean and fully counter balance. The legs are quite heavy compared to the upper body. Ideally I’d measure the torque in each joint, force on multiple points of each foot and like you said a gyro. But that is beyond the scope of this project.
Belevie it, it’s a very hard task. Just watch this video of the last Darpa challenge
https://www.youtube.com/watch?v=g0TaYhjpOfo
I can’t believe I haven’t seen this one before. This is pure comedy! (c:
A couple of them just folded back at the knees and fell, I wonder why? Power cut?
The last one was the best. I think it should’ve been arrested for DUI because it was obviously drunk. (c;
Btw, google just anounced that is working in a robot one order of magnitude better than Atlas.
https://www.youtube.com/watch?v=HyqT9Bdamt8&list=PL3H-OLcOPe066xrGbQTsIQc4sZaxR6bOr
Bad link.. This one is the good one.
https://www.youtube.com/watch?v=HyqT9Bdamt8
Why hasn’t anybody put a tail on a biped robot? Even birds have a tail for balance.
I can get bipeds to not quite walk (without a 60 dollar board), just as well as the one in the video, with minimal programming. Where is the video of the robot dancing that this control board was made to achieve? Sounds like a bad case of feature creep that turned into a kickstarter.
Makes me think that no one knows about Pololu’s selection of Servo Controllers, along with their scripting control software: https://www.pololu.com/category/12/rc-servo-controllers
David Buckley has done tons of work on Bipedal motion robots. Some of his minimalist designs use just 2 servos
http://davidbuckley.net/DB/A_Minimalist_Approach_to_Biped_Walking.htm
https://www.youtube.com/watch?t=13&v=upNFqsAYkBQ
What I’d like to see is someone taking a page from the Robotron and building a robot that uses a single motor to power all movement.
Single motor robots you say?
http://letsmakerobots.com/content/single-motor-robot-challenge
That controller and its software looks awfully similar to the good old Poloulu servo controllers: https://www.pololu.com/product/1356