Towards highly-tuned mobility in multiple domains with a dynamical legged platform.

ABSTRACT

As mobile robots become more commonly utilized in everyday applications, the tasks they are given will often require them to quickly traverse unprepared and varied environments. While traditional mobile platforms may falter under such conditions, animals utilize distinct locomotion modalities such as running, jumping, or climbing, to adroitly negotiate a wide variety of challenging and changing terrains. Due to limitations including available on-board power, legged robots have struggled to match the speed of these animals, even in a single mode of transport. In this paper we experimentally investigate the synergies and trade-offs in developing a dynamical legged robot capable of both running and climbing. We utilize bio-inspired 'templates' or reduced-order models of motion to identify how the dynamics change from running to climbing and seek to identify a minimal set of robotic adjustments necessary to switch locomotion modalities. This template-based design methodology is explained and the resultant robot behavior in each domain is characterized. We show that using a trotting gait, the platform demonstrates running speeds of up to 0.67 ms(-1) on level ground and climbing speeds of up to 0.43 ms(-1)on near-vertical surfaces (and up to 0.16 ms(-1) on vertical surfaces) while exhibiting dynamical behaviors comparable to that of the inspirational models.