ABM presentation Jun 8 final

18
Emergent Behaviours and their Simulations An introduction to Agent-based Modeling N Chandrasekhar Ramanujan @ IGCS June 8 2016

Transcript of ABM presentation Jun 8 final

Page 1: ABM presentation Jun 8 final

Emergent Behaviours and their Simulations

An introduction to Agent-based Modeling

N Chandrasekhar Ramanujan @ IGCS June 8 2016

Page 2: ABM presentation Jun 8 final

If you have seen a Mexican Wave you have seen an agent-based model at work.. • Link to mexican wave video• People in the stadium = the agents• Two simple rules : Stand when guy on left stands…. Sit when the guy

sits.

Page 3: ABM presentation Jun 8 final

Why do we make models?

• For insights • For explanations• To make predictions

Page 4: ABM presentation Jun 8 final

What does ABM do?• ABM helps us understand Complex Adaptive Systems• complex adaptive systems we refer to a group of (locally) interacting

agents, who act and react to actions of other agents.• The coherent emergent behavior that might occur in a system arises

from the local interactions of the agents• Unlike many other modeling techniques that quantify and then re-

create the patterns, agent-based models explore the causes of the patterns; the patterns are emergent properties from the individual decisions of the agents.

Page 5: ABM presentation Jun 8 final

What is an agent?• An agent is...

– An individual with a set of characteristics or attributes

• A set of rules governing agent behaviors or “decision-making” capability, protocols for communication

● Respond to the environment ● Interact with other agents in the system

• Agents are diverse and heterogeneous • This makes it interesting!

Page 6: ABM presentation Jun 8 final

Key elements of an ABM – the model• Identify the agents and define their actions• Agent consists of • Attributes• Behaviours – how do you parameterise these?

• Running the model – time is a discrete variable• Change conditions, tweak variables, see what happens!

Page 7: ABM presentation Jun 8 final

Emergent behaviour – flocking model in Netlogo• Introduce Net Logo – open source software• You will see a simulation of birds • They have been given three simple rules to control how they interact

with each other : essentially collision avoidance rules. • At the start of the simulation the birds are given random speeds and

directions. • However after a few moments the birds begin to fly in a formation

that is similar to the flocking formation seen in the real world.

Page 8: ABM presentation Jun 8 final

Agent-based models are now the tool of choice to model and simulate behaviours and outcomes in a number of contexts….

• Driving factors….• Huge Improvements in computing power • Huger improvements in data sets …. E.g. you can get real time data

on traffic for instance …. You can predict what-if scenarios and congestions

• Qn. : Where can we use here in Insti? Mess in the mess?

Page 9: ABM presentation Jun 8 final

Another example of ABM on Net Logo

• Take a simple wolf and sheep Model … and simulate small changes in rules• …. Simulate small changes in the environment… e.g. the fertility rates

of sheep affected by climate change….

Page 10: ABM presentation Jun 8 final

Why ABMs are preferred to general equilibrium, DSGE models etc

• Can use actual behaviours data and not assumptions of Rational Man behaviours • Coping with climate change …. How people may adopt different

behaviours response is not strictly determined by rational –economic coconsiderations

Page 11: ABM presentation Jun 8 final

Some Strengths of ABM• System assumptions• Heuristic ability• Heterogeneity• Bounded Rationality• Communication & Social Networking

Page 12: ABM presentation Jun 8 final

Some Shortcomings of ABMs • Data Problems: potential lack of adequate data due to nature of

model itself• Identifying rules of behaviours• Hard to prescribe solutions?

Page 13: ABM presentation Jun 8 final

ABM and GIS• GIS is a spatial modeling tool that stores and displays geographic data

and analyzes spatial relationships• Natural Synergy between ABM and GIS• Linking agent-based models to GIS means we can create models

directly related to space• Spatial data acts as a “container” for agents• Allows us to compare aggregate outputs to the “real” world, helping

us validate our models

Page 14: ABM presentation Jun 8 final

What can you do with ABM & GIS?• With ABM + GIS, you can:•Model and understand Land Use Change (Earlier Sriperumbudur

project by Keerthi, Amala)• Analyzing traffic congestion• Simulate spread of infectious diseases within populations

Page 15: ABM presentation Jun 8 final

Scenario planning and ABM• Planning is a very big, scary thing!• marked by complexity, multiple stakeholders and outcome uncertainty • ABM can help – how?• It is a method of a method that

• integrates a variety of data• allows for evaluation of comparable options• visualizes complex interactions and future outcomes

• develop thinking and expectations of how actors behave• see how linkages between actors affect behaviour

Page 16: ABM presentation Jun 8 final

Applying ABM in other contexts• In LOTR movies, where battles had 1000s of actors, used ABM with

special effects to model movements of people effectively• Budweiser ad – super bowl

Page 17: ABM presentation Jun 8 final

One for the Road…

• some fun assignments … how would you model these? What variables and behaviours would you take into account?• Crowd movement in saarang - • Traffic jam at Madhya Kailash• First day attendance at a new film release..• Stock of energy drinks at Gurunath …. Ebb and flow at different times of day/

week/ month / year….. Other Non temporal events like tests or quizzes

Page 18: ABM presentation Jun 8 final

Thanks for Listening! Welcome Questions, comments, suggestions