Omnidirectional Walking Robot

Spider robots have always interested me.  They have the potential of forming the basis of a highly maneuverable, flexible mobile platform.   And they look really cool!

However most of the current designs are not very fast (moving in a straight line) and can't carry much weight.  One of the main factors in this is the number of DOF each leg requires for basic mobility.  Most of these spider designs have three small motors per leg x 6 legs = 18 motors (or more).    Moreover the size and location of these motors limits the forces they can apply.  Even worse,  improving the fluidity of motion requires even more DOF per leg! 

On the other hand, the legged "animal" designs of Theo Jansen produce a very fluid organic motion with only a single motive force (wind or motor):

theo-jansen-2.jpg

Here's a simplified version from this Instructables page:

http://www.instructables.com/id/My-123D-Jansen-Strandbeest-Mechanism/

Leg-Assemly-Working-1-Animation(420)_4.gif

The inherent stability of this design comes from the fact that at any point in the cycle at least three legs are in contact with the ground at all times:

The only problem with using this design as a general robot platform is that the configuration of the linkages that drive the leg generally limit it to a linear motion. A few designs have attempted to make it more maneuverable by splitting it into two side-by-side units.  Each unit can be controlled independently, thus allowing for tank-like turning.  This works, but still does not provide a sufficient level of mobility for wide-ranging tasks.

I've thought about a number of approaches to adapt a Theo Jansen-esque design to an omnidirectional design.  Some approaches I explored included taking the basic leg pair and attaching it to a rotating joint, changing the linkages to make them pivot in 3 dimensions, and various other permutations.  Most of the designs resulted in an overly-complicated design with no inherent advantages over the existing spider layouts.

An alternate approach I've considered is to adapt a delta robot layout for the legs.  A delta robot uses three legs each driven by one motor kinematically coupled together to move the "foot" to an arbitrary point in 3D space.  

This design has the advantage of being inherently omnidirectional, and can be relatively easily adapted to a leg configuration by attaching the motors to each leg via worm gears:

To achieve a stable design would require at least 6 full leg assemblies like this.  This allows three legs to be in contact with the ground, while the other three legs are moving to the next location.  Taking the simplest design approach this would also require 18 motors, i.e. the same as a classic spider robot.  But even with 18 motors, using delta robots as legs provides considerably greater mobility, strength and smoothness of motion than an equivalent spider layout.

Moreover, I determined this approach could be further enhanced by coupling the 6 delta legs together into two sets of 3 legs each.  Coupling the legs this way means that the design can be reduced to six motors allowing the use of much larger, higher torque motors.  Each set of three feet can be moved to an arbitrary X/Y/Z position, and the center of gravity can be shifted simultaneously, so it can smoothly move in any direction.  I designed a series of six belts connecting the sets of motors together.  Three belts are placed on one level and three on another so that they don't intersect:

My next step is to build a prototype.  I'm going to 3D print the plates and legs and use Nema 17 motors.  I've broken the model into thirds so that I can print a third of it first to test-fit everything together:

I'll post updates as I make progress on the testing.

 

 

 

 

 

 

 

 

 









Posted on June 7, 2014 .