Game theory Lecture 19. Dynamic games. Game theory

Similar documents
MS&E 246: Lecture 4 Mixed strategies. Ramesh Johari January 18, 2007

Long-Run versus Short-Run Player

6.254 : Game Theory with Engineering Applications Lecture 8: Supermodular and Potential Games

Mixed Nash Equilibria

An Introduction to Noncooperative Games

Noncooperative continuous-time Markov games

Extensive Form Games I

Game Theory. Greg Plaxton Theory in Programming Practice, Spring 2004 Department of Computer Science University of Texas at Austin

Economics 209A Theory and Application of Non-Cooperative Games (Fall 2013) Extensive games with perfect information OR6and7,FT3,4and11

Political Economy of Institutions and Development: Problem Set 1. Due Date: Thursday, February 23, in class.

Mean-field equilibrium: An approximation approach for large dynamic games

Distributed Learning based on Entropy-Driven Game Dynamics

Economics 201B Economic Theory (Spring 2017) Bargaining. Topics: the axiomatic approach (OR 15) and the strategic approach (OR 7).

Basics of Game Theory

EC319 Economic Theory and Its Applications, Part II: Lecture 2

Near-Potential Games: Geometry and Dynamics

Non-zero-sum Game and Nash Equilibarium

Quitting games - An Example

Near-Potential Games: Geometry and Dynamics

Stability of Feedback Solutions for Infinite Horizon Noncooperative Differential Games

Computing Minmax; Dominance

Copyright (C) 2013 David K. Levine This document is an open textbook; you can redistribute it and/or modify it under the terms of the Creative

Computing Equilibria of Repeated And Dynamic Games

EC3224 Autumn Lecture #04 Mixed-Strategy Equilibrium

3.3.3 Illustration: Infinitely repeated Cournot duopoly.

Repeated Downsian Electoral Competition

Microeconomics. 2. Game Theory

Exercise 11. Solution. EVOLUTION AND THE THEORY OF GAMES (Spring 2009) EXERCISES Show that every 2 2 matrix game with payoff matrix

Backwards Induction. Extensive-Form Representation. Backwards Induction (cont ) The player 2 s optimization problem in the second stage

6.254 : Game Theory with Engineering Applications Lecture 7: Supermodular Games

Computing Solution Concepts of Normal-Form Games. Song Chong EE, KAIST

Theory Field Examination Game Theory (209A) Jan Question 1 (duopoly games with imperfect information)

Evolution & Learning in Games

Dynamic Games, II: Bargaining

6.891 Games, Decision, and Computation February 5, Lecture 2

Advanced Microeconomics

Game Theory. 4. Algorithms. Bernhard Nebel and Robert Mattmüller. May 2nd, Albert-Ludwigs-Universität Freiburg

Computing Minmax; Dominance

FPTAS for Computing a Symmetric Leontief Competitive Economy Equilibrium

Game Theory. Professor Peter Cramton Economics 300

1 Motivation. Game Theory. 2 Linear Programming. Motivation. 4. Algorithms. Bernhard Nebel and Robert Mattmüller May 15th, 2017

For general queries, contact

Game Theory DR. ÖZGÜR GÜRERK UNIVERSITY OF ERFURT WINTER TERM 2012/13. Strict and nonstrict equilibria

Coalitional solutions in differential games

Outline for today. Stat155 Game Theory Lecture 17: Correlated equilibria and the price of anarchy. Correlated equilibrium. A driving example.

A Price-Based Approach for Controlling Networked Distributed Energy Resources

Computational Game Theory Spring Semester, 2005/6. Lecturer: Yishay Mansour Scribe: Ilan Cohen, Natan Rubin, Ophir Bleiberg*

Games of Elimination

Selecting Efficient Correlated Equilibria Through Distributed Learning. Jason R. Marden

Game Theory Fall 2003

Bayesian-Nash Equilibrium

Title: The Castle on the Hill. Author: David K. Levine. Department of Economics UCLA. Los Angeles, CA phone/fax

Lecture 1. Evolution of Market Concentration

6.254 : Game Theory with Engineering Applications Lecture 13: Extensive Form Games

REPEATED GAMES. Jörgen Weibull. April 13, 2010

EC3224 Autumn Lecture #03 Applications of Nash Equilibrium

Computation of Efficient Nash Equilibria for experimental economic games

Evolutionary Stochastic Games

MS&E 246: Lecture 17 Network routing. Ramesh Johari

Weak Dominance and Never Best Responses

האוניברסיטה העברית בירושלים

Notes on Blackwell s Comparison of Experiments Tilman Börgers, June 29, 2009

BELIEFS & EVOLUTIONARY GAME THEORY

Spring 2016 Network Science. Solution of Quiz I

EC319 Economic Theory and Its Applications, Part II: Lecture 7

Bounded Rationality, Strategy Simplification, and Equilibrium

Belief-based Learning

Negotiation: Strategic Approach

Optimal Control Feedback Nash in The Scalar Infinite Non-cooperative Dynamic Game with Discount Factor

Lecture 6: April 25, 2006

Game Theory, Evolutionary Dynamics, and Multi-Agent Learning. Prof. Nicola Gatti

Finding Optimal Strategies for Influencing Social Networks in Two Player Games. MAJ Nick Howard, USMA Dr. Steve Kolitz, Draper Labs Itai Ashlagi, MIT

What kind of bonus point system makes the rugby teams more offensive?

Global Dynamics in Repeated Games with Additively Separable Payoffs

Ü B U N G S A U F G A B E N. S p i e l t h e o r i e

Game Theory and its Applications to Networks - Part I: Strict Competition

Game Theory and Algorithms Lecture 7: PPAD and Fixed-Point Theorems

Game Theory Review Questions

Game interactions and dynamics on networked populations

On Equilibria of Distributed Message-Passing Games

Random Extensive Form Games and its Application to Bargaining

Game Theory and Control

Monotonic ɛ-equilibria in strongly symmetric games

Equilibria in Games with Weak Payoff Externalities

SF2972 Game Theory Exam with Solutions March 15, 2013

Game Theory and Algorithms Lecture 2: Nash Equilibria and Examples

Convex Sets Strict Separation. in the Minimax Theorem

Introduction to Game Theory

A Few Games and Geometric Insights

Advertising and Promotion in a Marketing Channel

arxiv: v1 [math.oc] 7 Dec 2018

CS 573: Algorithmic Game Theory Lecture date: January 23rd, 2008

Brown s Original Fictitious Play

Extinction in common property resource models: an analytically tractable example

Evolutionary Game Theory

Econ 618: Correlated Equilibrium

CS364A: Algorithmic Game Theory Lecture #13: Potential Games; A Hierarchy of Equilibria

Games with Perfect Information

MS&E 246: Lecture 12 Static games of incomplete information. Ramesh Johari

Introduction to game theory LECTURE 1

Transcription:

Lecture 9. Dynamic games

. Introduction Definition. A dynamic game is a game Γ =< N, x, {U i } n i=, {H i } n i= >, where N = {, 2,..., n} denotes the set of players, x (t) = f (x, u,..., u n, t), x(0) = x 0, x = (x,..., x m ), 0 t T, indicates a controlled system in the space R m, U,..., U n are the strategy sets of players,..., n, respectively, and a function H i (u,..., u n ) specifies the payoff of player i N.

. Introduction A controlled system is considered on a time interval [0, T ] (finite or infinite). Player strategies represent some functions u i = u i (t), i =,..., n. Depending on selected strategies, each player receives the payoff H i (u,..., u n ) = T 0 g i (x(t), u (t),..., u n (t), t)dt+g i (x(t ), T ), i =,..., n Actually, it consists of the integral component and the terminal component; g i and G i, i =,..., n are given functions. There exist cooperative and noncooperative dynamic games. Solutions of noncooperative games are comprehended in the sense of Nash equilibria.

. Introduction Definition 2. A Nash equilibrium in the game Γ is a set of strategies (u,..., u n) such that H i (u i, u i ) H i (u ) for arbitrary strategies u i, i =,..., n.

2. Discrete-time Dynamic Games (Fish Wars) Imagine a certain dynamic system governed by the system of difference equations x t+ = (x t ) α, t = 0,,..., where 0 < α. For instance, this is the evolution of some fish population. The initial state x 0 of the system is given. The system admits the stationary state x =. If x 0 >, the population diminishes approaching infinitely the limit state x =. In the case of x 0 <, the population spreads out with the same asymptotic line.

2. Discrete-time Dynamic Games (Fish Wars) Suppose that two countries (players) perform fishing; they aim at maximizing the income (the amount of fish) on some time interval. The utility function of each player depends on the amount of fish u caught by this player and takes the form log(u). The discounting coefficients δ (player ) and δ 2 (player 2) are given, 0 < δ i <, i =, 2. Find a Nash equilibrium in this game.

First, we study this problem on a finite interval. Take the one-shot model. Assume that players decide to catch the amounts u and u 2 at the initial instant (here u + u 2 x 0 ). At the next instant t =, the population of fish has the size x = (x 0 u u 2 ) α. The game finishes and, by the agreement, players divide the remaining amount of fish equally. The payoff of player makes up H (u, u 2 ) = log u + δ log ( 2 (x u u 2 ) α) = = log u + αδ log(x u u 2 ) δ log 2, x = x 0. Similarly, player 2 obtains the payoff H 2 (u, u 2 ) = log u 2 + αδ 2 log(x u u 2 ) δ 2 log 2, x = x 0.

The functions H (u, u 2 ) and H 2 (u, u 2 ) are convex, and a Nash equilibrium exists. For its evaluation, solve the system of equations H / u = 0, H 2 / u 2 = 0, or αδ = 0, u x u u 2 And so, the equilibrium is defined by u = u 2 αδ 2 ( + αδ )( + αδ 2 ) x, u 2 = αδ 2 x u u 2 = 0. The size of the population after fishing constitutes x u u 2 = αδ ( + αδ )( + αδ 2 ) x. α 2 δ δ 2 ( + αδ )( + αδ 2 ) x. The players payoffs in the equilibrium become H (u, u 2) = (+αδ ) log x +a, H 2 (u, u 2) = (+αδ 2 ) log x +a 2,

where the constants a, a 2 follow from the expressions ( αδ j (α 2 δ δ 2 ) αδ ) i a i = log [( + αδ )( + αδ 2 ) ] +αδ δ i i log 2, i, j =, 2, i j. Now, suppose that the game includes two shots, i.e., players can perform fishing twice. We have determined the optimal behavior and payoffs of both players at the last shot. Hence, the equilibrium in the two-shot model results from maximization of the new payoff functions H 2 (u, u 2 ) = log u + αδ ( + αδ ) log(x u u 2 ) + δ a, x = x 0, H 2 2 (u, u 2 ) = log u + αδ 2 ( + αδ 2 ) log(x u u 2 ) + δ 2 a 2, x = x 0. The Nash equilibrium appears from the system of equations αδ ( + αδ ) = 0, u x u u 2 αδ 2( + αδ 2 ) = 0. u 2 x u u 2

Again, it possesses the linear form: u 2 = αδ 2 ( + αδ 2 ) ( + αδ )( + αδ 2 ) x, u2 2 = αδ ( + αδ ( + αδ )( + αδ 2 ) x. We continue such construction procedure to arrive at the following conclusion. In the n-shot fish war game, the optimal strategies of players are defined by u n = αδ 2 n n (αδ ) j (αδ 2 ) j x, u n (αδ 2 ) j 2 n = j= αδ n n (αδ ) j (αδ ) j x. n (αδ 2 ) j j= ()

After shot n, the population of fish has the size x u n u n 2 = α 2 n δ δ 2 (αδ ) j n (αδ 2 ) j n (αδ ) j n (αδ 2 ) j j= As n, the expressions (), (2) admit the limits u = lim n un = Therefore, αδ 2 ( αδ )x ( αδ )( αδ 2 ), u 2 = x. (2) αδ ( αδ 2 )x ( αδ )( αδ 2 ). where x u u 2 = lim n x un u n 2 = kx, k = α 2 δ δ 2 x ( αδ )( αδ 2 ).

Suppose that at each shot players adhere to the strategies u, u 2. Starting from the initial state x 0, the system evolves according to the law x t+ = (x t u (x t ) u 2(x t )) α = k α x α t = k α (kx α t ) α = k α+α2 x α2 t =... = k t j= α j x αt 0, t = 0,, 2,... Under large t, the system approaches the stationary state x = ( ) α α αδ +. (3) αδ 2 In the case of δ = δ 2 = δ, the stationary state has the form ( ) α x = αδ α 2 αδ.

We focus on the special linear case. Here the population of fish demonstrates the following dynamics: x t+ = r(x t u u 2 ), r >. Apply the same line of reasoning as before to get the optimal strategies: u n = (δ 2 ) j δ 2 n x, u n n (δ ) j (δ 2 ) j 2 n = j= (δ ) j δ n x. n n (δ ) j (δ 2 ) j j= As n, we obtain the limit strategies u = δ 2 ( δ )x ( δ )( δ 2 ), u 2 = δ ( δ 2 )x ( δ )( δ 2 ).

As far as x u u2 x = δ + δ 2, the optimal strategies of the players lead to the population dynamics ( ) t r x t = δ + δ 2 x r t = δ + x 0, t = 0,,... δ 2 Obviously, the population dynamics in the equilibrium essentially depends on the coefficient r/( δ + δ 2 ). The latter being smaller than, the population degenerates; if this coefficient exceeds, the population grows infinitely. And finally, under strict equality to, the population possesses the stable size. In the case of identical discounting coefficients (δ = δ 2 = δ), further development or extinction of the population depends on the sign of δ(r + ) 2.