Fuzzy Logic Expert System for TierScalable Planetary Reconnaissance

1

• Fuzzy Logic Expert System for Tier-Scalable
  Planetary Reconnaissance
• Roberto Furfaro ,James Dohm, Wolfgang Fink
• Aerospace and Mechanical Engineering Dept
• Department of Hydrology and Water Resources,
• Lunar and Planetary Laboratory
• University of Arizona
• Visual and Autonomous Exploration Systems
  Research Laboratory,
• California Institute of Technology
• SpaceOps 2006, June 19-23, Rome

2
Tier-Scalable Mission Architecture Concept
Paradigm-shift in space exploration Tier-scalab
le mission architecture for unconstrained,
science-driven robotic missions (Fink
and Dohm PSS 2005)
3
Tier-Scalable Mission Architecture Round-Robin I
4
Tier-Scalable Mission Architecture Round-Robin I
5
Autonomy for Tier-Scalable Missions

• Unconstrained, Science-Driven Planetary
  Reconnaissance requires higher level of on-board
  automation
• Autonomous determination of sites with the
  highest probability of significant scientific
  findings and/or natural resources
• Solution Fuzzy Expert System for tier-scalable
  autonomy
• Collects information at multiple scales using
  multiple instruments mounted on multiple
  platforms
• Synergistically connected to AGFA-like smart
  software
• Performs synthesis of spatial and temporal
  information
• Exhibits high degree of flexibility can be tuned
  to autonomously infer presence of life, water,
  and any other natural resources

6
Expert System Basic Structure
7
Fuzzy Logic Fundamentals

• Fuzzy Logic (FL) is a multi-valued logic that
  allows intermediate values to be defined between
  conventional evaluations such as yes/no,
  hot/cold, high/low
• FL is an alternative to traditional two-valued
  Greek (Aristotelian) logic
• FL is about relative importance of precision
• Precision is not Truth Henry Matisse
• FL is a convenient way to map an input space
  into an output space
• FL has been used in multiple applications
• Control systems of industrial processes,
  Automatic classification, Household electronics,

8
Why Fuzzy Logic ?

• FL is conceptually easy to understand
• The concepts behind the successful implementation
  of a fuzzy expert are natural and very simple
• FL is flexible
• Modularity can be easily implemented in a fuzzy
  logic framework. Rules can be add on top of
  existing knowledge-base
• FL is tolerant of imprecise data
• Everything is imprecise if we look close enough.
  Fuzzy logic builds on the process rather than on
  the precision
• FL can be built of top of experience of experts
• Expertise built over years by planetary
  scientists, geologists, astrobiologists etc. can
  be easily incorporated into the system
• FL is based on natural language
• The basis for fuzzy logic is the basis for human
  communication

9
Fuzzy Sets and Membership Functions

• Fuzzy sets describe vague concepts
• Hydrogen is high, Chlorine is low
• Fuzzy sets admits partial membership
• The degree of belonging to a fuzzy set is
  described by the Membership Function (MF)
• MF maps an input (crisp) value into a membership
  value

10
Logical Operations with Fuzzy Sets
Fuzzy Sets A, B
Fuzzy AND
Fuzzy NOT
Fuzzy OR
11
Knowledge-Base and Inference Mechanism (I)

• The basic element of the knowledge-base are
  IF-THEN rules.
• E.G.IF hydrogen is High and Chlorine is High
  THEN Potential for Water containing is High
  (Confidence factor 1)
• Two basic parts Antecedent and Consequent
• Three basic steps are required to perform fuzzy
  interpretation of the rules
• Fuzzification of the inputs via the membership
  functions
• Application of the fuzzy operators to the
  antecedent
• Application of the implication method
• The forth step, de-fuzzification, is required to
  obtain a crisp value

12
Knowledge-Base and Inference Mechanism (II)
Step 1 Fuzzify the inputs
Step 2 Apply OR operator
Step 3 Apply implication
Step 4 De-fuzzification (e.g. centroid method)
13
Following-the-water Fuzzy Expert System
Architecture
14
Building knowledge base via synthesis-of-
information approach
Approach Geologic detective work through
comparative analysis of layers of published
information to unfold the history of Mars
Collectively, evidences/traits add credence to
working hypotheses and overarching Theories and
help address anomalies in the hydro- geologic
history of Mars
Odyssey (e.g., Elemental GRS) and Mars Express
15
Water on Mars (Kargel et al.)

• It exists now most abundantly as ice in polar
  caps and in permafrost at middle latitudes, with
  mineralogically bound water in the tropics
  (present water).
• We know from geology that water was once abundant
  as surface rivers, lakes, and seas at lower
  latitudes large glacial ice sheets once spread
  from the poles and mountain ranges to lower
  latitudes and elevations. The observations
  indicate geologically brief but important
  climatic forays to wetter, warmer conditions
  (Past Water).

16
Polygons Signs of Permafrost Freeze-Thaw?

• GEOLOGY WAS OUR TIP THAT MARS IS AN ICY WATER
  WORLD
• For example
• Polygons occur across Martian middle and
  circumpolar latitudes.
• Here we see cracks developed on the floor of an
  impact crater at a high southern latitude.
• Their location correlates globally with neutron
  signatures indicating abundant hydrogen (ice) in
  the upper meter.
• On Earth, similar polygons form through the
  action of thermal contraction cracking and
  repetitive freeze-thaw of ice-rich permafrost.
• Other formative mechanisms should be considered
  mud desiccation and salt-wedge hydration/dehydrati
  on.

17
Recent and Past Water-related Activities
18
Following-the-water Water Indicators

• Define the indicators that quantify past and
  extant water
• Evidence of present water can quantify by
  hydrogen content
• Evidence of past water includes elemental,
  spectral, geomorphological, stratigraphic and
  topographic indicators

19
Following-the-water Fuzzy Rules
20
Fuzzy System Design and Implementation

• System has been designed to Follow-the-Water
• Seven Input parameters/indicators and one output
• Membership functions for the input/output system
  have been defined
• Conventional MIN/MAX construct have been used for
  AND/OR operators
• The MIN implication method has been used as
  inference technique
• Aggregation executed via summation of all
  possible outputs
• Centroid method applied to defuzzify output
• System coded and implemented using MATLAB Fuzzy
  logic Toolbox

21
Membership Functions for input/output parameters
22
Testing the system Tier-Scalable Simulation
Simulated Scenario Tier-scalable reconnaissance
over Solis Planum, Thaumasia region (Mars)
Multi-sensor data maps
23
Fuzzy System Simulation Results
24
Fuzzy Inference Results Highest Probability
Locale
Potential For Water Containing 79.9
Chlorine 0.9
Sulfates 20
Hydrogen 9
Hematite 18
Sapping Channels 9
Valley Networks 3
Basins 8
25
Fuzzy Inference Results Lowest Probability Locale
Potential For Water Containing 20.64
Chlorine 0.2
Sulfates 4
Hydrogen 1
Hematite 2
Sapping Channels 1
Valley Networks 1
Basins 0
26
Conclusions and future effort

• Autonomy is required for successful tier-scalable
  missions
• The fuzzy expert system is based on planetary
  science expertise that can be implemented using
  the natural language
• Design, implementation and testing show that the
  system is effective in following-the-water
• The system is modular and flexible. It can be
  modified to incorporate multidisciplinary
  knowledge
• Current and future efforts include looking for
  sites where life could flourish as well as
  looking for sites that contain/hide natural
  resources on extraterrestrial planetary bodies
• Tier-Scalable on the Moon Fuzzy expert system to
  detect areas with high probability of finding
  water