Abstract: | Research into legged robots predates the discovery of the neural basis of rhythmical motion in animals by several decades. As a result almost no research in legged robotics has ever been based on this, and indeed most is still derived from personal insight or previous biological theories which are not known to be flawed. One group which has used these ideas to design a control structure for a robot is Taga and his colleagues, who made a crude simulation of a planar biped controlled by neural oscillators. in this paper we replicate and extend Taga's work, correcting flaws in the original model which made it extremely mechanically implausible. disappointingly, it was found that the original model was only stable in a very limited range of conditions. However, making the model more realistic has improved this considerably, and the authors are now investigating further similar biologically inspired control structures.
|