Comparing Effectiveness of Bioinspired Approaches to Search and Rescue Scenarios
Emily Shaeffer and Shena Cao
4/28/2011Shaeffer and Cao- ESE 313
4/28/2011Shaeffer and Cao- ESE 313
Combine: • The Ant Colony Optimization (ACO)
convergence mechanism• Bees Colony task division-forager, scout,
packers• Cockroach Swarm Optimization automatic
swarming=
• Efficient navigation in 2D discrete environment between home base and target "danger" locations, faster than these algorithms alone
C3.4 Hypothesis
4/28/2011Shaeffer and Cao- ESE 313
C3.1 Desired Behavior or Capability: Swarming for Improved Search and Rescue
• What is Swarming? Large groups to accomplish large tasks Algorithms for ants, bees, cockroaches
• Use of Swarming for Search and Rescue “Foraging Task”- Can be performed by robots
independently, multiple improve performance Sept 11- robots found nothing, swarming robots
could have covered more ground Focus on searching and mapping, not rubble
removal or extraction• Why Swarming
Collective intelligence for non-intelligent robots
4/28/2011Shaeffer and Cao- ESE 313
C3.2 Present Unavailability: Where Robots are Lacking
• Current Technology Separate algorithms modeling the behavior of
each type of insect Using just the cooperative collaboration model
of ants, improved navigating Ability to change between tasks increases
efficiency
• Missing Technology A combination of all three techniques for most
efficient possible navigation in different scenarios
4/28/2011Shaeffer and Cao- ESE 313
C3.3 Desirability of Bioinspiration: 3 Different Insect Inspired Algorithms
• Ant colony optimization algorithm Ants go any direction, pheromone trail strength
indicates shortest path Used Pure ACO
• Artificial bee colony Higher efficiency by task division using
foragers, scouts, and packers BeeSensor Routing
• Cockroach Swarming Chase-swarming behavior, dispersing behavior,
ruthless behavior
4/28/2011Shaeffer and Cao- ESE 313
Combine: • The Ant Colony Optimization (ACO)
convergence mechanism• Bees Colony task division-forager, scout,
packers• Cockroach Swarm Optimization automatic
swarming=
• Efficient navigation in 2D discrete environment between home base and target "danger" locations, faster than these algorithms alone
C3.4 Hypothesis
4/28/2011Shaeffer and Cao- ESE 313
• Create Basic Obstacle Grido GridWorld
2D environment Bounded Discrete Provided:
Actor class-random movements which interact with other actors
Flower objects that decay over time (humans or pheromone trail)
Station rocks that can interact (change colors-might mark what has been found)
• Test refutability parameters
C3.6 Necessary Means
4/28/2011Shaeffer and Cao- ESE 313
• Detection time-found all danger zones on map
• % Humans saved in time
• Behavior judged relative to 3 algorithms alone
C3.5 Refutability
4/28/2011Shaeffer and Cao- ESE 313
4/28/2011Shaeffer and Cao- ESE 313
• Created grid implementations in which all actors could interact with each other
• Each test scenario contained at least one victim, obstacles, and different combinations of other actors
• Have scenarios for only ants, only bees, and only cockroaches
Results: Grid Implementation
4/28/2011Shaeffer and Cao- ESE 313
• Cockroach Swarm Optimization• Set visibility range (90 degree angle in
forward direction)• Find local best (calculate individuals
proximity to object and find closest)• Move randomly towards local best• Local best reaches target, marks it and
moves to next target• If clustered, individuals interact and
increases probability of dispersion (from 0.1 to 0.5)
• Values yet to be optimized• Have yet to implement other algorithms
• Vision: using the pure ACO concept on the path of bee colony algorithm
Detailed Implementation
4/28/2011Shaeffer and Cao- ESE 313
• Cockroach Swarm Optimization• Performs well for dispersing and moving
between target sites• Speed?
• ACO• Good speed• Search?
• BeeSensor• Good combining factor
• Therefore we still believe that our final implementation will surpass these algorithms individually
Predicted Results
4/28/2011Shaeffer and Cao- ESE 313
• Understanding• More thorough understanding of weaknesses in
literature• Understanding of implications of weaknesses
in literature• Further defining what optimization is and what
the literature considered optimization• More mathematical analysis to better predict
what our results would be even if the code is not working
Next Steps
4/28/2011Shaeffer and Cao- ESE 313
• Need more time to work though code so we can test our different scenarios
Conclusions
2/28/2011Shaeffer and Cao- ESE 313
Questions?
2/28/2011Shaeffer and Cao- ESE 313
Supplementary Slides
2/28/2011Shaeffer and Cao- ESE 313
1) Randomly disperse from base, find food
2) Randomly retract back to base, leave pheromone trail
3) Step proportionate evaporation of pheromone trail
4) Probabilistic following of pheromone trail
5) Positive feedback leads to optimization
Ant Colony Optimization Details
2/28/2011Shaeffer and Cao- ESE 313
1) Start with base
2) Each bee finds neighboring source, respond with “wiggle dance” based on nectar amount
3) Onlookers evaluate response, changesources accordingly
4) Best sources found
5) Positive Feedback Effect
Artificial Bee Colony Details
2/28/2011Shaeffer and Cao- ESE 313
1) Chase-Swarming behavior Each individual X(i) will chase individual P(i) within its visual scope or global individual Pg
2) Dispersing behavior At intervals of certain time, each individual may disperse randomly X ′(i) = X (i) + rand(1, D),i = 1,2,..., N 3) Ruthless behavior Current best replaces an individual selected at random X (k)=Pg
Cockroach Swarming Details
Reference: Chen ZH, Tang HY (2010) 2nd International Conference on Computer Engineering and Technology. 6, 652-5