Hexapod Gaits - and a CHALLENGE

[rpcook]

Thought I’d start a discussion topic about the sort of movement that we might be able to get out of the Hexy kits when we get our hands on them.

I’m interested in the locomotion of the particular gaits that one can use with a hexapod (there’s enough degrees of freedom for several).

Quick search of the net suggests the following three basic hexapod movements:
[ul][]Tripod Gait - legs operate as two independent tripods[/]
[]Wave Gait - legs operate one at a time; slow, but stable[/]
[]Ripple Gait - a wave moves along each side of legs[/][/ul]
I’m sure that we’ll get at least one of these implemented out of the box, but I’d love to hear/ see what others might be able to come up with. What about “stotting” http://en.wikipedia.org/wiki/Stotting? Does the Hexy have the power in its servos to do that? Something really novel perhaps? Criteria is to achieve linear motion, not necessarily forwards.

What about gaits for “injured” Hexys (NB please do not hurt your Hexy ). Hexapods in nature (e.g. insects) can adapt their gaits after injury or losing a limb. What can we mimic with a stock Hexy?

How would one judge the “goodness” of these gaits? I’ve read a paper about horse locomotion that correlated the energy consumption of the horse (via oxygen usage in breath) to the gait at a particular speed. Seemingly nature has gotten very good at not expending too much energy so the animal will change gait as the relative efficiencies of the available gaits switch over. Could we produce some standardised method for the Hexy kit?

As a challenge to the community, I propose the following: a Hexy Gallop. E.g. Hexy at a full running pace, with all six legs off the ground for some period during the stride. Can this be done? Is the hardware even going to be up to this level of power output? Does control of this high-speed gait require more sensory input than the Hexy comes with out-of-the-box? I can offer the following prizes:
[ul][]Confirmed gallop with a stock Hexy: 10 internet points![/]
[]Confirmed gallop with a modified Hexy: 5 internet points![/][/ul]
I suspect that one might need an accelerometer to feedback how long to place each stride; perhaps a Hexy carring a smartphone will be sufficient for this challenge?

[georgefrick]

This is very interesting, but as I am still working on a single leg; I’m not quite to working on a gait. All my other robots are wheeled or tracked.

I hope to be able to work towards meeting one of these challenges.

[mkean]

So one of the biggest challenges I foresee is using servos to generate a sprung-leg effect. Most forms of running require the leg to be used as a spring during contact, harnessing energy at landing and releasing it leading to liftoff. The elasticity of the leg is what makes the gait efficient. Servos won’t have the elasticity or the ability to store and release energy like that. Having said that, you SHOULD be able to slightly modify the legs to include a spring that would allow this type of energy harnessing. I don’t know enough about servos to know off hand if letting them freewheel during this motion would damage them or not. So I see it being possible with minor modification to your hexy kit, but not ideal out of the box.

As for sensory information necessary for these gaits, identifying touchdown and liftoff should be sufficient if you have elasticity in the leg. There are many provably stable running control algorithms that require no specific knowledge other than that. Look up the SLIP model and work by Mark Raibert for the control algorithms. They’re from the mid 80’s and early 90’s. He’s got a book about hopping robots if you’re interested in it.

I’m very interested to see a wave gait on Hexy myself. I’m personally hoping to be able to run it using a home-made central pattern generating neural network. You have an interconnected network of simple signal generators (neurons). They are attached to each other so that they either help others activate or stop them from activating. Some of them send signals to the motors when they activate. The interconnections cause a pattern to be generated as they activate each other and the servos. The gait transitions you mention in nature actually come from these types of pattern generators sending nerve impulses to the muscles. At low speeds they generate one pattern and as the speeds increase their connections automatically transition gaits from the slow moving one up into the faster moving one. Very interesting but really complex stuff.

EDIT:
As long as your hexy has 4 “uninjured” legs you can resort to an ambling gait. The trick would be choosing which four legs to use and how to use them for straight line motion if they are asymmetrically arranged about HEXY. Simple if a pair of legs goes down (MR and ML for instance) as you just ignore the paired limbs that don’t work and act like a quadruped walking along. It’d be fun to see someone transitioning between ambling and paired tripod walking on the fly.