Running up a wall: the role and challenges of dynamic climbing in enhancing multi-modal legged systems

Abstract

Animals have demonstrated the ability to move through, across and over some of the most daunting environments on earth. This versatility and adaptability stems from their capacity to alter their locomotion dynamics and employ disparate locomotion modalities to suit the terrain at hand. As with modalities such as running, flying and swimming, dynamic climbing is commonly employed by legged animals, allowing for rapid and robust locomotion on vertical and near-vertical substrates. While recent robotic platforms have proven effective at anchoring reduced-order, dynamic climbing models, its adoption as a common modality for multi-modal, legged platforms remains nascent. In this work, we explore several of the open questions related to the physical implementation of dynamic climbing, including investigation of substrate inclinations for which dynamic climbing is suited, mitigation of destabilizing out-of-plane dynamics and improvement of attachment reliability in the presence of dynamic effects. The results from these inquiries provide several mechanisms and approaches for increasing the reliability and versatility of dynamic climbing as a dynamic legged modality. With these and other developments into legged locomotion modalities, future multi-modal platforms will begin to approach the expertise of biological creatures at moving through a complex and challenging world.

AquaClimber: a limbed swimming and climbing robot based on reduced order models

“…Additional parameter sweeps revealed that incorporating springs with properly tuned stiffness can produce a 20% gain in speed and (with an added lateral leg sprawl offset angle) stability improvements. Though unconventional for swimming, these match findings for our climbing robots [9]. Further, when comparing the model's two modes, it predicted the trade-offs in speed and compliance efficacy between swimming and climbing.…”

“…Freely employing each of these methods as the domain demands differ, enables animals to move seamlessly in even the most challenging environments such as shear mountains, dense rain forests, ocean cliff-sides, underwater caverns, etc. Individually, this breadth of behaviors has been reproduced by a variety of legged robots some of which can run [1][2][3], jump [4][5][6], or climb vertical surfaces [7][8][9]. Combining these sets of these behaviors onto singular platforms which can match animal-like modal flexibility, however, is difficult.…”

Tracked Wall-Climbing Robot for Calibration of Large Vertical Metal Tanks

“…The suction/propulsion adhesion generates a pressure difference for climbing on the smooth surface [8,9]. The mechanical grasping adhesion utilizes grasping mechanisms like a bio-inspired spine or tiny hook for climbing [10,11]. The chemical adhesion makes use of adhesive tapes on the robot feet to climb on the vertical surface [12].…”

Impact of slope on dynamics of running and climbing

“…The SLIP model inspired platforms able to approach animal-like velocities, including: Raibert's hoppers, which were first platforms to achieve stable running locomotion with froude numbers greater than 1 (which formally differentiates fast walking from running) [11], iSprawl able to move at 15 body lengths per second (BL s) [12], and MIT's Cheetah II (with a mass of 33 kg) able to move at 9.1 BL s −1 (6.4 m s −1 ) [13]. Additionally, the Full-Goldman model has inspired the fastest legged climbing robots which have reached velocities of 60 cm s −1 [10] and 1.95 BL s −1 [14].…”

“…In the horizontal domain, animals like the cockroach use their individual leg pairs differently, but it is unclear how the leg forces within individual leg pairs transition from pushing to pulling (discrete or continuous transition) as the incline changes, and what implications these transitions may have for the associated reduced order models. While past modeling work extended the LLS model to study pushing or pulling based locomotion on slopes [8] and previous experimental work with a FG-based climbing robot examined the running performance as slope was decreased from vertical to horizontal [14], the fundamental LLS or FG models must be extended to directly examine the dynamics as slope is changed. Therefore, new models are required to explore the mechanisms by which the dynamics change, whether it be from changes in gait, frequency, or individual leg function.…”