A General Rhythmic Pattern Generation Architecture for Legged Locomotion - Advancing Artificial Intelligence Through Biological Process Applications - page 217

j r =

2

r = = = ; r r 1

Figure 8. Two equivalent architectures to illustrate the SMER algorithm, where i m n

r = . Left, the original SMER graph description. Right, an alternative description for the composite
OBB module with the macroneurons and their clones.

3

i

i

n

n

j m

m

j

k

k

Table 1. Pseudo-code operation of a macroneuron

/*Program: Macroneuron_firing_function*/
CONST:
r[i : 1…m] : integer;
threshold[i, j : 1…m], weight[i, j : 1…m] : real;

VAR:
membrane[i, j : 1…m] : real ;
firing[i] : boolean;

Initial State:
membrane[i, j] := random;
firing[i] := false;

Action at a local pulse:
if ( ∃ j ∈ ( m ) ( j ! = i ) ∧ ( weight [ i , j ]! = 0 ∧ ( membrane [ i , j ] < threshold [ i , j ]))

begin
firing[i] := false;
await NOT(membrane[i,j] < threshold[i,j]);
continue;
end
else
begin
firing[i] := true;
if ( ∀ j ∈ ( m ) ( j ! = i ) ∧ ( weight [ i , j ]! = 0 ))

begin
membrane[i,j] := membrane[i,j] - weight[i,j];
membrane[j,i] := membrane[j,i] + weight[i,j];
continue;
end
end
end