Breaking the Kubernetes Kill Chain: Host Path Mount
Livestock Evacuation Plan
1. Creating an evacuation plan for livestock
during emergency disasters: The case of
Fukushima, Japan
By:
Chrysafis Vogiatzis, Ruriko Yoshida, Ines Aviles-Spadoni, and Shigeki Imamoto
2. March 11, 2011
• Magnitude 9.0 - 70 km east of Oshika Peninsula
– 15,790 deaths, 5,933 injured 4,056 missing across 18
prefectures, 125,000 buildings damaged or destroyed
• Most powerful earthquake to ever hit Japan - one of five most
powerful in the world
• 40 m high tsunami - 10 km inland.
• Meltdown at three reactors in Fukushima
Nuclear Power Plant complex
• 20 km radius evacuation zone
3.
4. Why we decided to get to work on a plan?
• Economic implications to region, farmers
• Loss of lively hood
• Public health implications
• Humane implications
5.
6.
7. What kind of assistance is out there?
•After Hurricane Katrina in 2005, the Pets Evacuation and
Transportation Standards Act (PETS Act, Pub. Law No. 109–
308) was passed into law in 2006
•The Federal Emergency Management Agency does have
information for dealing with livestock in a variety of
disasters but none formulated mathematically
•And, no specific evacuation plans for livestock in areas
where nuclear power plants are located
8. So we set our sights on…
• Providing the livestock industry a tool to protect their animals
and farm, in this case, those located near nuclear power
plants
• IFAW Report states “It is essential to devise a plan to save the
greatest number of animals possible in the shortest amount
of time through the work of these specialized teams.”
• Our efforts
• Completely altruistic
• Non-funded
9. What does this mean
for the U.S.?
• Total of 104
reactors
• 20% electricity
• Eastern half of U.S.
• 52 are 40-years-old
• Beef, 32,834,801
• Dairy, 9,266,574
10.
11.
12. Prior to civilian
evacuation, the Japanese
Ministry of Economy,
Trade and Industry
reported that, 9300
head of cattle were in
the Fukushima
Prefecture including
30,000 pigs and 440,000
chickens
13. Our Research
• We came up with two new models for
livestock evacuation
• We present also an algorithm based on the
augmented Lagrange relaxation
• Computational results and comparisons are
also given to depict the success of our
approach.
14. Why use OR?
• What is Operations Research?
– A discipline that employs mathematical
formulations and techniques to model and solve
real life problems.
• Why use it?
– Provides us with the best possible solution
– Makes planning and management to problems
more efficient
– Has had an impact on social welfare (aviations,
logistics, scheduling)
15. Evacuation Management
• It is a special, large scale optimization problem
• Techniques suitable for smaller instances
might prove to be computationally costly
– Evacuation modeling is a special case of a network
design problem
– NP-hard; i.e. difficult to solve
– Approximation schemes are used to reduce
running time
16. Models
• Mathematical formulations need to be tight
and rigorous
• Different models assess different measures
and approach other solutions
• We show two models:
– Origin-destination problem
– Network design evacuation problem
17. O-D Evacuation Problem
Assumptions:
– Time horizon until end of evacuation phase
– Simulation results for danger progress
– Accurate livestock numbers
maximizing total outflow
flow preservation
capacity
standard success probability
nonnegativity
18. O-D Evacuation Model
• Time Dynamic model, which implies a time
extended network
• Flow preservation constraints for time
dynamic models make our LP solution non
integral (loss of total unimodularity)
• Stochastic optimization involved
20. Network Design Evacuation
• NP-hard
• VRP-based formulation
– |K| vehicles that have to evacuate animals
minimize total cost
percentage constraint
all percentages sum <=1
vehicle capacity
0 <= each percentage <= 1
22. Lagrange Relaxation
• Barrier method: exterior point sequences
• This implies our solution might be non integral
in the end
• However,
– easier problem to solve (getting rid of the coupling
constraint)
– in the time given we can solve it to approximate
an integral solution
27. Computational Results
• O-D Formulation:
– Maximum optimality gap noted: 2.9%
• Average gap of 1.4%, which can be practically ignored!
– Speedup achieved up to 180%
• Network Design Formulation:
– Maximum optimality gap noted: 22.8%
• Average gap of just 5.6%!
– Speedup achieved up to 600%
28. Future work
• Create and test more models
• Come up with a better algorithm, i.e. a
decomposition scheme
• Working with prefectures, counties to
compare with actual real life plans
(technology transfer)
Good afternoon everyone, my name is Ines Aviles-Spadoni. I am from the University of Florida and I am also a member of the Hachiko Coalition, a group that was formed after the the March 11 , 2011 triple disaster in Japan.I am here today to talk to you about an evacuation plan for livestock during emergency disasters, specifically the case of Fukushima, Japan.The path to creating this plan was done using mathematical techniques/formulations that were created by doctoral student ChrysafisVogiatzis, of the Optimization Center in the department of Industrial and systems engineering at the University of Florida and overseen by Dr. Ruriko Yoshida, of the University of Kentucky.
After seeing report after report of the loss of livestock in the area, we thought we could be proactive by using our skills to come up with an evacuation plan. Our hope is by creating such a plan, we can save animals lives, distress to farmer (as these animals are their livelihood and many do grow close to their herds, and save the industry money as well. In the case of Fukushima, the authorities simply did not prepare for anything, so this is a worst case example.Several papers have been written about what to do when handling externally contaminated livestock, and many of them advise against mass euthanasia to avoid a disruption in the food chain until the Lethal Dose at which they have been exposed is deemed acceptable.
When the devastating earthquake occurred on March 11, 2011 followed by a 30 foot tsunami that hit the eastern coast of Japan, it severely damaged the Fukushima Daiichi power plant, prompting authorities to order an evacuation of all humans living within 20 km of the nuclear plant. People were ordered to go to evacuation shelters, they were told they could not bring their pets with them but they could return in 3 days apparently there were no plans to evacuate livestock to safer grounds. People were told by the Japanese authorities that they could come back in three days, but days passed and then the zone became restricted.Animals, pets and livestock were left to perish…livestock secured in their pens with enough food and water for three to four days as well as countless dogs, cats and other pets as we can see from the images.
However, the PETS law does not account for livestock.Fema has information for dealing with livestock but no evacuation model that has been formulated mathematically Furthermore, no mathematically formulated plan for livestock evacuation.
We also chose to created an evacuation model to give farmers or the livestock industry a tool to protect and save their lively hood, especially those located close to nuclear plantsLost of livestock during any disaster economically and emotionally devastating for farmers.In many cases, farms/animals represent generations of hard work, improving their stockLoss of farms mean loss income for them and the regionMany farmers grow close to their animals, even the ones to be sold as meatHaving the expertise to understand a problem via mathematical methods, to model for evacuation planning, and animal sciences background, my collaborators and I thought to put our mental resources together to come up with an evacuation plan to give prefectures, counties, states, animal rescuers and farmer a tool for preventing a disaster such as this one in the future.
The reactor with the highest risk rating is 24 miles north of New York City, in the village of Buchanan, N.Y., at the Indian Point Energy Center. There, on the east bank of the Hudson, Indian Point nuclear reactor No. 3 has the highest risk of earthquake damage in the country. 20% of the world’s nuclear power plants are in earthquake zones
That’s a lot of livestock in this regionAnd due to the tragedy, there were obvious losses to the cattle, swine and poultry industry.Definite financial losses to the livestock industry. This could have been prevented by having an evacuation plan in place.
Evacuation after a football match (Gators): efficiently and effectively move people in and out of the stadium – no mass confusion, chaos etcNorth East Florida Evacuation Planning (Jacksonville) for hurricanes: hurricane-prone, port city, that has numerous plans for different scenariosEmergencyManagement is using Operations Research techniques and ideas
We have techniques readily available (commercial solvers etc) that can tackle small instances-problems. The transportation network is HUGE even in rural areas. So, evacuation becomes an extremely large and hence hard problem to solve. Evacuation NEEDS TO BE timely. That’s the reason why we apply our formulations and algorithms to it. Algorithm: input -> algorithm -> output a smart idea of repeating certain tasks until you are satisfied with the resultComputationally costly: timely, costs too much time to computeNetwork Design Problem: decide which roads to useNP-hard: small instances up to 40-50 nodes can be solved efficiently, however for 100 nodes there is NO solution. It could take multiple days to show infeasibility (that it cannot be solved)!!Approximation scheme: algorithm that does not necessarily find the best solution. But, in a very small time, it can find a “good” solution.
Mathematical Formulation: a series of equations that describe a physical problem. Each of the equations is like a rule that needs to be obeyed. \\Tight: the smallest possible set of “rules” (aka equations) that need to be satisfied -> then the time it would take to solve is smallerDifferent models describe the problem in a diverse way (others are more realistic but extremely difficult while others are unrealistic but serve as good initial solutions)O-D problems: we have a set of demands (animals, livestock and so on) at known places and we have a set of safety points (destinations). Each of the demands wants to reach safety, subject to a series of restrictionsNetwork Design Problem: choose which roads to use, basically it is like “building” the networkThe next two models describe the problems of livestock evacuation with different assumptions.
Time horizon means for example that I have a limited time to get animals out (after that the evacuation is stopped)Simulation results will give us an idea of HOW the disaster progresses for each given time. That shows us which areas become inaccessible, which roads can and cannot be used after a certain time. Simulation is not ALWAYS true: I am only taking the probabilities. In the end, what I get from the simulation is a series of probabilities that a road can still be used for each road in the network. Getting all livestock out, less realistic.Maximizing total outflow: maximize the number of animals that leave the dangered zone and enter safety at every timeFlow preservation: If I have x animals arriving to a certain place, then x animals will leave that place in the futureCapacity constraint: Every road has limited capabilities and can only serve a certain number of trucksStandard success probability: We have a “limit” in the probability of success, meaning that if the roads that the plan says to use are most probably dangerous, I will not use those. The limit is set at 99%, but can be selected. Nonnegativity: every physical element must come in positive numbers (I cannot have -3 animals or -2 trucks)Stochastic model: depending on probabilitesSimulation provides us with road status probabilitiesSelection of routes with high probability
VRP: Vehicle routing problem: notoriously difficult problem to solve. More realistic, more difficult to solve.|K| vehicles: I have a set number of trucks that I can use to evacuate animals – more realistic approachpercentage constraint: the percentage of animals saved is the animals picked up in total divided by the animals that were originally thereAll percentages sum up to less than or equal to 1: I cannot save more animals than there are (if 100 cows need to arrive to safety, I cannot save 110 of those)Vehicle capacity: If my vehicle can save up to 50 cows, I cannot put 60 on it…assumption is set number of vehicles that can go in and save the animals,but it makes them more difficult to solve.(12): if a road is used by a vehicle, then it “exists” for use from our model
Important to understand what the algorithm has done. Commerical solvers take too long to solve these two models (days). Instead of looking for best possible solutions, its best to look for a good solution that is perfect but that can be computed or solved faster (in a smaller time period). These are the technical details of what the algorithm does.Description: Lagrange relaxation, you take out some of the constraints or rules, and you resolve an easer problem. As soon as you are done, it is not guaranteed to have that specific rule/constrained satisfied. In order to ensure all rules are met, we round up the final solution.
Solves a simplified version of the original problem (I.e. taking out some of the “rules”) and then resolving until we arrive at a better solution. After that, we round all fractional values up. Lagrange relaxation, which is a relaxation of the “rule set”.All computational results were taken from the Fukushima network. Different networks were created for a different level of detail (granularity). It is much like zooming in on a map: first few nodes are considered and then as I zoom in, more roads appear and other areas as well. What does this mean for emergency mangagement or livestock evacuation? That we have proved that our algorithm provides competivite results with the currently used methods to figure out an evacuation plan (such as commercial solvers C-Plex and Gurobi), it solves the problem in less time with a little more expensive (for examples, it is more expensive and may save a fewer less animals or use more resources, trucks), but quickly gives us a plan. The other commercially used solvers talk longer to use and are more better results (optimal), but during a unforseen disaster such as fukushima, then we need something rather faster and good than slower and perfect.With this information, prefectures, counties, states, would now be able to layout plans for livestock evacuation in a much quicker way if they choose to use this method ranther than their commercial solvers.
Every effort must be made to immediately treat specified animals currently in the restricted area(companion animals, agricultural animals and wild animals anywhere in the entire prefecture of Fukushima). IFAW reportGroups involved in animal rescue operations must closely coordinate and establish the effectivecommunication and cooperation required for this work. Maintaining human safety as first priority,rescuers must have adequate knowledge, training and ability with respect to protection againstradiation, as well as be equipped with dosimeters to measure radiation levels. It is essential todevise a plan to save the greatest number of animals possible in the shortest amount of timethrough the work of these specialized teams