SIGMA robot gait improvements

A new chassis, a more permanent (if still rather ugly) circuit board -- but most importantly, a major bug fix in the gait calculations. (The "knee" angle calculations didn't take into account the "shoulder" angle changes. This resulted in a very inefficient gait.)

Ideally, I would like to get a physics simulation of the robot running and connected to a genetic algorithm program, to evolve an optimal gait. For now, though, it's effective enough.

Next up: gradual turns, turns-in-place, and reversing. After that, a second microcontroller will be added, to provide higher-level control (most likely, programmed in C, since it will be handling more complex functions. The servo routines are written in assembly, since they need every clock cycle they can get.

Robot plans to be made available on www.paleotechnologist.net/SIGMA .

Category:

Science & Technology

Tags:

• robot
• quadruped
• servo
• PIC
• Drexel
• Goodwin
• paleotechnology
• paleotechnologist
• FlyByPC

[skyhacker6]

You cheat on the way of quadrupède walking method. Your robot is just craxling on the floor because the feet are always in contact with the floor ! See my method whith 9 servo on : robot quadrupède by skyhacker62.

[FlyByPC]

@skyhacker6 The feet do actually come up off the floor a bit; this is possible with two servos per foot. The gait is still not optimal, but it does walk instead of shuffle.

[emrobot]

Hi - I'm wondering about the gait - can you confirm that it's moving one foot/leg at a time?

lol - which joint is "knee" and which is "shoulder"? - you're mixing appendages!!

It's neat how you've arranged the CofG to move in concert with the moving stability footprint (uh, I can't think of other words to describe it.. I mean the triangular shape formed by the three feet (on the ground), where the CofG must project onto to preserve stabiility?!)

[FlyByPC]

By shoulder, I mean the joint at the base of the leg. The phase of the gait was actually inspired by watching a video I took of a horse walking. (I figured Nature had a few million years to figure it out...) The path that each foot takes was basically guesswork, trying to match what made sense. I'd like to actually evolve a gait via genetic algorithms, once I get a physics sim running.

[rantaoca]

lol

AUTONOMY

i was waiting for it to turn or something =D

and it just rammed into the wall!

but all in all, hella great of a robot.

good job!

[FlyByPC]

Maybe -- there are no onboard sensors, for now; all of the movements are preprogrammed. Wireless control will probably be next, and then I will look into autonomy. (It may have to move to a larger platform, first.)