Comparing different controllers for the coordination of a six-legged walker

I tuned out a little during this paper. He is comparing three/four controllers of coordination: (Cruse 98), (Porta, Celaya 2001), and (Steinkuhler and Cruse 1998)

The problem of this paper is in different methods of "Determining the positions of AEPs and PEPs, i.e. deciding when to lift legs and where to put them". This is the third of his list of the general subproblems of controlling the gait of a six legged walker. What is remarkable to me, and i didn't realize it until reading this paper, is that as much disagreement as there seems to be about models, and as many approaches as there are, all the papers I have read have conceptualized the problem in a way that this framing can describe.

Defintions:

• "Gaits are referred to as 'fixed' if internally produced recurring patterns are used for coordination."
• "..the free-gait group = coordination is generated by local rules, which may lead to somewhat irregular patterns."

Importantly: "Free gaits have the advantage that they are more flexible and can in principle cope with more difficult situations. On the downside , the generation of free gaits is an accordingly more difficult problem because it is not obvious how to guarantee body stability in every situation." What I wrote in the margin was '!analyst fetish!". Is it possible to guarantee body stability in every situation? Is it possible with a fixed gait in an unpredictable environment? This leads to another framing of what is becoming a central question for me. What procedure can one follow, and what metrics will prove useful, in analyzing distributed systems with the goal of providing confidence that a given distributed system is superior to or inferior to a given centralized system for a given problem in a somewhat given environment. Looking at the above sentence, it obviously won't give full predictability or any guarantees. It won't be able to look at every situation. What determines a satisfactory tool is a social metric. What is 'assurance enough' for a community comfortable with centralized systems in simplified environments and accustomed to having solutions to sets of differential equations?

More: "Advantages of local rules are their superficial simplicity and - resulting from that - their simple implementation. The behavior of a system of local rules can, however, only be predicted with difficulty. Therefore, it is also hard to design such a system to meet certain criteria (e.g. body stability)."

right after that: "Systems of local rules commonly generate free gaits." I like the separation here of local rules from free gaits, it reminds me of local rules that can generate fixed gaits, though the separation isn't consistent with the above definition, which suggests causation of the latter by the former.

I didn't put the sentence by sentence attention necessary to fully appreciate all the conclusions of this paper. Roughly, after testing stability during many tests (curve walking, inclines and declines, etc) the two variations on the Porta model came clearly out on top. I think it was also the simplest of the three. Apparently, the original was intended for high v and, in this application, had to be modified to suit the platform, which was slower.