-- MarkCutkosky - 24 Jul 2004

Biology Design Lessons

Current state of what I think we're learning from observations of animals that climb vertical surfaces well, and what the implications are for design and control of our robots' feet.

1. Insects and lizards that excel on hard, non-smooth vertical surfaces invariably use spines or claws in addition to any adhesive methods.

We need to use both spines and adhesives too. Initially, development of each capability is taking place independently, to simplify testing & optimization (e.g., reduce chance of confounding). In the next phase, integrated feet with both adhesion and spines will be developed.%ENDCOMMENT%

2. On touchdown, insects and geckos produce a short sliding motion in the stroke direction to

• reliably engage multiple spines
• preload the adhesive pads

and thereby obtain maximum traction and adhesion.

In experiments with compliant multi-spined feet, we also have noticed that a short sliding motion in the stroke direction allows multiple spines to find useful contacts – provided that the spines are substantially decoupled from each other while the sliding takes place. Then, the spines should be “locked up” or rendered less compliant as the foot starts to bear substantial load in shear. Preloading, with or without sliding, also seems to be important for maximizing the available pull from adhesives. %ENDCOMMENT%

3. Spine or claw effectiveness on hard surfaces is a direct function of spine size versus surface roughness. (See data and paper by Zani; see fractal scaling argument by Cutkosky).

This observation has lead us to use larger numbers of smaller spines for better adhesion and traction on hard exterior surfaces such as concrete and stucco. (Concete is very rough to a cockroach, not so rough to a cat). See Zani’s diagram of a claw tip pulled against a hemispherical bump for sizing. However, this many-spined approach only works if a large percentage of the spines can be assured of finding asperities to hold on to. Otherwise, either the spines or the surface material will fail when loaded. This, in turn, leads us to create feet in which the spines can be decoupled from each other -- through compliance or an articulation or a sliding degree of freedom – while the preloading action, mentioned above, takes place. %ENDCOMMENT%

4. There is a characteristic pattern of:

• pull-in in the normal direction for the front limbs (to prevent pitch-back)
• outward normal force at the rear limbs to maximize traction.
• lateral pull-in at all limbs to increase stabilization.

Guessed but not confirmed: In part, this force pattern reduces swinging should one of the feet suffer a transient failure.

The lateral forces appear to be helpful when climbing with spined feet. We have observed in our experiments with the multi-tendon mult-spined prototypes that it is important to keep the machine from swaying or swinging when one foot loses its grip. This is important to prevent a transient failure from becoming a catastrophic one. %ENDCOMMENT%

5. On convex surfaces, inward lateral forces (and inward-directed internal forces at the contacts more generally) enhance traction and pull-in. Conversely, for concave surfaces the internal forces should be outward-directed, as they are in running.

The convexity/concavity argument is easily established through static or quasi-static analysis. We do not yet have a machine that can regulate the internal forces effectively to test this in practice. We also do not know for certain whether animals revert to outward directed internal forces (as in running) when climbing concave surfaces. %ENDCOMMENT%

6. Geckos (and perhaps insects?) exhibit a construction in which one or more tendons is routed to distal claws or pads, with multiple attachment points. This is a useful way to control the distribution of shear stresses over the surface or over the collection of spines. Without something like the tendons, what tends to happen is that shear stresses build up at the edge of the contact area that is in the direction of the pulling force, causing it to disengage or peel off prematurely.

For our robots, adapting this principle has been instrumental to obtaining success on difficult vertical surfaces. Otherwise, they are highly vulnerable to undesired disengaging or unpeeling. %ENDCOMMENT%

7. Lizards and insects use active attachment and detachment strategies. They don't rely entirely on passive mechanisms. With active spine or pad engagement, the chances of getting a good grip on the wall are enanced. Why? I think the reason is that one can customize the motion produce by the distal active engagement to facilitate engagement of spines or preloading of adhesive pads in a way that might be difficult to do with the main leg motors. The problem is that the main leg actuators are the result of some design compromises (work-space kinematics, obtaining adequate forces at the feet, velocity requirements, etc.) and are not optimized for engaging spines or pads.

We initially resisted adding extra, distal degrees of freedom, but I am beginning to think they will be necessary for success of challenging vertical surfaces. In truth, the added complexity and weight are less than one might imagine. The actuators can be low-powered (because the movements and velocities are low) and can perhaps even be binary (up/down or in/out) rather than proportional. Using low friction push-pull cables as tendons, the actuators can be located on the robot body. %ENDCOMMENT%