Walking is a means of locomotion that is ubiquitous among terrestrial animals and the matter of considerable technical inquiry; both for biological understanding and description, and engineering construction and control. Although wheels and treads have numerous advantages over legs for low-complexity terrain, the promise of adept legged locomotion in a much broader range of rugged environments is eloquently demonstrated in the animal kingdom. Of primary interest in the understanding of such agility is the ability of animals to smoothly transition between behaviors requiring substantially different local behavior of locomotor appendages.
Recent developments in our understanding of insect walking systems, encapsulated in the study of the neural mechanisms of stick insect leg coordination by (Ekeberg, Blümel, & Büschges, 2004), have made it possible to construct models of local leg control based on known properties of biological systems. Such models can provide the appropriate “ports” to investigate and predict the effects of descending commands in the transition between and generation of different local behaviors.
This dissertation describes the development and use of robotic models of step generation to address questions about descending control. Robotic models were desired both for the ease of experimental interaction and the fidelity of physical modeling they can provide. The NeuRoMod software suite was developed, and provides interactive operation and experimental scripting for the robotic models.
The local control methods of the stick insect described by Ekeberg et al. are standardized as Sensory Coupled Action Switching Modules (SCASM), and tools for the use of this concept in modeling are developed and demonstrated. The apparent generality of SCASM as a computationally simple control concept is also addressed.
Experiments were conducted to demonstrate model usage, in the testing and generation of biologically relevant hypotheses. The basic function, neuromechanical nature, and resilience of SCASM-controlled steppers is demonstrated. Simple muscle models are found to provide significantly improved, reliable stepping. Control methods for forward walking and inside turning in the cockroach model are presented, and experiments are conducted investigating how descending influences can cause transitions between these behaviors. The results support the general reflex cascade hypothesis presented by (Mu & Ritzmann, 2008a).