C4: Project Formulation

wonderfuldistinctΤεχνίτη Νοημοσύνη και Ρομποτική

16 Οκτ 2013 (πριν από 4 χρόνια και 2 μήνες)

70 εμφανίσεις


ESE 313

February 29, 2011


Adam
Komoroski

Carol Wong


C4: Project Formulation

Overview

Problem Statement:


1. Desired Behavior:


2. Present Unavailability:


3. Desirability of Bio
-
inspiration:

The Hypothesis


4. The Idea


5. Refutability


6. Necessary Means





C.1 : Desired Capability


Observed Problem: complete loss of one leg function


Observed Result: unpredicted and uncontrollable motions


Desired Behavior: fault tolerance; directed, controlled, and
purposeful movement.


Ability to compensate for the complete loss of function in one
leg, restoring motion to a state comparable to full legged
functions.


Bio
-
inspiration:


Variations in leg load bearing


Rat motor cortex injuries


Neural plasticity


C.2 Present Unavailability

"The problem of fault recovery represents a vast, important domain in
its own right that is still relatively unexplored in robotics.”
[2]

-
Much research done on fault
-
tolerable gaits in the specific
instance of locked joint failure

-
Does not apply to Junior’s locomotive system

-
Ideas can be extrapolated

C.3 Desirability of Bio
-
inspiration


Mathematical models available


Biological observations that are relevant made



Rats with motor cortex injuries:


Solid observations made that can be extrapolated to Junior
platform


Plasticity :Animal analogs have uncanny ability to adapt to
injuries and other sustained handicaps


Load Bearing


Remaining questions:


How do we implement an artificial rendering of neural plasticity?



C.4 The Idea


Current implementation:
http://kodlab.seas.upenn.edu/Aaron/Iros10
, :43


Five legged “crawl”


All five legs offset from each other


Upper left leg (0) loss of function


3,1,5,2,4


Four Buehler clock parameters same across all 5 legs


Transition to stable gait


Proposed implementation:


Focus: purposeful disabling of one corner leg


Goal: transition effectively, stabilize resulting crawl gait


Optimize velocity of five legged gait


Stabilizing transition


Mathematical approach: optimize parameters via
Nelder

-
Mead and
machine learning algorithms


Bio
-
inspired approach: vary load bearings on select legs

C.5 Refutability


Evaluation of Performance:


Stable gait: readings of IMU, acceleration, center of mass
changes


Efficient: energy expenditure


Directed gait: ability to transverse pre
-
designated path;
observation


Controlled gait: velocity controlled and optimized; velocity
tracker



C.6 Necessary Means


Proposition:

1.
Bioinspired

approach:

-
Determine parameters that change load bearing on a given leg

-
Purposefully influence load bearing characteristics on legs

-
Evaluate performance

2.
Mathematical model approach:

-
Nelder

Mead Algorithm: optimization of ‘objective function’

-
Objective function: characterizes system parameters and behavior

-
Machine learning classifiers

-
Collect data
-
> observe behavior
-
> formulate model
-
> model =
objective function




QUESTIONS?

Five Legged “Crawl” Gait

Nelder

Mead Algorithm