Why our Perching Proposal should fit in CPP: MEDIUM: ...

* Perching is an example of a system that requires integrated cybernetic (sensing, control, computation) and physical solutions -- tightly integrated. The reasons are:

  • dynamic maneuvers with short time constants. control authority approaches zero as you approach the wall, necessitating open-loop strategies and tuned physical systems to take over in the final stages of landing.
    • The objective of the physical system is to (i) absorb kinetic energy (ii) stabilize the plane (e.g., keep pitch, roll angles in desired range) and prevent bouncing (iii) direct forces toward the feet to faciliate engagement with the surface.
    • The objective of the cybernetic system is to (i) identify suitable surfaces for landing using information from sensors & other planes (ii) choose a landing strategy and corresponding robust control for the initial landing stages (iii) to monitor the progress of the plane while under control and after feeback control becomes ineffective to ensure that the plane is approaching a set of initial conditions such that a successful landing will occur. (iv) If the probability of failure becomes too high the cybernetic system must take the plane into an ``abort'' maneuver such that damage is avoided and the plane and recover and try again (v) record variables related to plane performance, local ambient conditions (wind, surface conditions) for analysis and learning to improve subsequent landings by this and other planes.

Cybernetic problems:

Surface identification: initially using optical sensing + using context (e.g. GPS and knowledge of surfaces in the vicinity) and pattern recognition. Once a plane has contacted the surface, there is much more accurate information regarding surface characteristics (roughness, skewness, kurtosis) important for grasping. The spines used to grasp the surface are also styli, similar to those used in standard profilometers. This detailed information is used to update the estimates from optical sensing.

Robust control:

Learning of trajectories for different ambient conditions:

Physical problems:

Analysis and synthesis of nonlinear compliant, damped mechanisms to stabilize the plane and facilitate grasping the surface for a wide range of initial conditions (initial pitch, roll angles, initial linear and angular velocities on contact). This is a constrained mechanism optimization problem. The degrees of freedom include the topology of them mechanism (how many links, what connections) and the parameters of the mechanism (dimensions, stiffness and damping in links and joints). A nonlinear suspension is required as the requirements are substantially different at initial surface contact versus when the plane is coming to rest. Constraints include: avoid bouncing off the surface (this is the most common failure mode); avoid having undesired components prematurely contacting the surface (e.g. the "knee" joint of the mechanism or the tail of the plane); avoid excessive forces that would damage spines and/or the surface (this is especially important on relatively fragile surfaces like stucco and plaster).

-- MarkCutkosky - 29 Dec 2009

 
This site is powered by the TWiki collaboration platformCopyright &© by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback