Lessons learned from recent very large-scale disasters in the world
1. Ripple-Spreading Models and
Algorithms for Integrated Risk
Governance
(Some preliminary results)
Dr Xiaobing Hu
State Key Laboratory of Earth Surface
Processes and Resource Ecology
Beijing Normal University
(Collaborating with University of Warwick)
2. Outline of talk:
• Motivation
• Ripple-spreading models
• Ripple-spreading algorithms
• Conclusions and future work
5. Natural ripple-spreading phenomenon:
• Ripple-spreading
phenomenon or effect
widely exists in nature.
• The development or
evolvement of many
complex systems is largely
determined by the spreading
influence of a few local
events (e.g., disasters).
Does it reflect certain organization principle?
6. Ripple-spreading network model:
X.B. Hu, M. Wang, et. al., Physical Review E, Vol.83, No.4, pp. 046123, 2011.
Pool
Sensors
distributed in the
pool
Stone thrown
into the pool
7. Ripple-spreading network model for IRG:
• Ripple spreading The spreading of
impact/influence of disasters
• Node activation threshold Node resilience
• Links between nodes Disaster chain
• Node amplifying factor Cascading effect
• Final topology Damage assessment
8. Merits of ripple-spreading network model:
• Temporal factors + spatial factors.
• Great flexibility and freedom in modification
(e.g., deterministic model to stochastic model,
constructive model to destructive model, energy
feedback feature and multiple activation feature).
• The deterministic feature makes it easy for
parameter tuning, and even makes it possible to
optimize the network topology.
10. Application to epidemic model:
J.Q. Liao, X.B. Hu, M. Wang, et. al., CISP-BMEI 2012, Nov 2012, China.
Simulation result of the SARS outbreak in
HongKong in 2003
11. Application to scheduling model:
X.B. Hu, and E. Di Paolo, Evolutionary Computation, Vol.9, No.1, pp.77-106,
2011.
13. Optimization principle of ripple-spreading:
A ripple spreads at the same speed in all directions, so it
always reaches the closest node first in the space.
14. Ideal for route optimization:
• Tested in
resource/facil
ity location
optimization
problem,
which is an
important
issue in risk
management.
• Nearly 100
times faster
than existing
algorithms.
15. For multi-objective optimization:
X.B. Hu, M. Wang, et. al., IEEE Transactions on Man, Systems and Cybernetics ,
accepted, 2012.
• Theoretically
and practically
capable of
calculating the
complete true
Pareto front.
• New methods
for decision-
making in risk
management,
which is obviously
of multi-objective.