Game Theory Game theory is a branch of mathematics. It was first devised by John Von Neumann. It provides tools for predicting what may happen when stakeholders with conflicting interests interact. . Slide 1 Introduction Before we comment anymore on game theory lets consider a soccer game. A group of players playing against another group to score goal(s) Do all the players on a group have the same strategies? The individual strategies of each player converges to attain the objective of group.(why?) How is soccer game different from a two player game like squash? . Slide 2 Three basic components: Players Strategies Payoff (preference relationship) Let‘s now think of very basic players, strategies and payoffs in telecommuncation. Players (Nodes, Users, Operators, New/Handover calls) Strategies (modulation scheme, amount of bandwidth, transmit power etc) Payoff( Revenue, QoS, Call admission, lower BER etc) 01.01.2022 . Slide 3 Game Theory Does the decision of one player affect the decision of other? Let‘s consider a game, where a man(buyer) with limited budget is to buy some groceries at a grocery store, the decision of setting the price dervies the selection of grocery item, meaning thereby payoff of one player is dependent on the payoff of other player….. Can you think of any such scenario in mobile services…. Suggest players and their payoffs? 01.01.2022 . Slide 4 Analogy of example in Telecom Provider – 4 Provider – 3 Provider – 2 Provider – 1 User Pool Service Area 01.01.2022 Both Providers & Users can be modelled as buyers / Grocery store. . Slide 5 Consider the figure, if you are to start your journey at point „s“ and terminate at „t“ and you are provided with two routes. What are the parameters that you evaluate to choose the one of the two routes? 1. Distance 2. Others using the same route Cost Function Think of a similar scenario in Wireless Communication… 01.01.2022 . Slide 6 A simple Game theory example - It is a two player game (row palyer and column player) - Player-1 chooses the row and player-2 chooses column - The values in each cell represent utilities of players - First number in the cell is utility of player-1 - Second number in the cell is utility of player-2 Prisonner‘s Dilemma C = Cooperator – dont testify D = Defect – testify Let‘s now mathematically define a strategic game…………. 01.01.2022 . Slide 7 Formal Game Definition Normal form (strategic) game a finite set N of players a set strategies Si for each player i N ui (s) payoff function for each player i N where s S jN S j is the set of strategies chosen by all players s S is a set of strategies chosen by players ui : S 01.01.2022 . Slide 8 Common Games Zero Sum Games They are true games of conflict, Any gain of my side comes at the expense Of my opponents. E.g. Matching pennies game. Player-1 gets a Euro from Player-2 if both choose the same strategy or otherwise loses a Euro Battle of Sexes A couple wants to spend evening together, wife(P-1) wants to go to Opera and husband(P-2) wants to go football match (1,-1) (-1,1) (-1,1) (1,-1) (2 , 1) (0,0) (0,0) (1,2) Normal form game is one instance of repeated game played between large populations of P-1s & P-2s 01.01.2022 . Slide 9 Dominated Strategy Let‘s consider P-2 • Is M better than R? • yes (R-dominated) • No (M-dominated) Knowing this P-1 will .. M • B dominated? What do we observe? M R 4,3 5,1 6,2 2,1 8,4 3,6 3,0 9,6 2,8 T • T dominated? • M dominated? L B Do we have a unique strategy profile that both players agree to play? 01.01.2022 . Slide 10 Nash Equilibrium In the last slide we observed that Neither player has a unilateral incentive to change its strategy (Nash Equilibrium) In any strategic game given by G ( N , Si , ui ) A strategy profile is a Nash Equilibrim, such that for s*there S exists every player iN ui (s *i , s *i ) ui (s *i , si ) Few Natural Questions: 1. Do Nash Equilibrium always Exist? 2. Are Nash Equilibrium unique? 01.01.2022 . Slide 11 Solving the Game (min-max algorithm) Player 2 Player 1 A B C D A 4 -10 7 0 7 B 3 2 5 8 8 C 2 0 1 -4 2 D 5 -1 3 -5 5 2 -10 1 -5 choose minimum entry in each choose maximum entry in each row column choose the maximum among choose the minimum among these these this is maximin value this is the minimax value if minimax == maximin, then this is the Nash point of game 01.01.2022 . Slide 12 Multiple Nash Equilibriums In general, game can have multiple saddle points Player 2 A Player 1 B C D A 3 B 2 C 2 D 5 2 5 8 8 -10 2 0 2 0 2 -4 2 -1 3 -5 5 2 -10 2 -5 Same payoff in every Nash strategy unique value of the game Strategies are interchangeable Example: strategies (A, B) and (C, C) are Nash Strategies then (A, C) and (C, B) are also Nash Strategies 01.01.2022 . Slide 13 Cooperative Games In cooperative games coalitions are formed among the players and all the players then strive to increase the payoff of coalition. Coalition represents an agreement between players in the set coalition. The Coalition value in quantifies the worth of coalition in a game. A coalition game is defined as ( N , v ) The most common form of coalition game is characteristic form, whereby the value of coalition depends on members of that coalition with no dependence how the players of set other than coaltion is structured. The characteristic function of coalition quantifies the gain of S. The characteristic function of empty coalition is zero and satisfy the superadditive property. v( S ) v(T ) v( S´U ´T ) 01.01.2022 . Slide 14 Core The solution to coalition games is core Given a grand coalition N, a payoff verctor x R for dividing v(N ) is a group rational if x v( N ) . A vector is individually rational if every player can obtain a benefit no less than acting alone i.e. xi v({i}) An imputation is payoff vector satisfying the above two conditions. Thus core is defined as iN i cTu {x : xi v( N ), xi v( S )} iN iS Go through TU, NTU cooperative games 01.01.2022 . Slide 15 Bargaing Problems Bargaining problems refer to the negotiation process (which is modeled using game theory tools) to resolve the conflict that occurs when there are more than one course of actions for all the players in a situation, where players involved in the games may try to resolve the conflict by committing themselves voluntarily to a course of action that is beneficial to all of them. 01.01.2022 . Slide 16 Definition & Axioms Bargaining problem is modelled as as pair (F; d), where F represents theset of all feasible utility pairs and d is the disagreement point. Players will not form coalition if the utility that they receive is lesser than disagreement point. The most common solutions that exist for bargaining solutions include Nash Bargaining solutions, Kalai-smorodinsky bargaining solutions etc. All such solutions have to satisfy few axioms namely i) individual rationality ii) pareto optimality iii) independence of irrelevent alternative / individual monotonicity iv) Symmetry. 01.01.2022 . Slide 17 Application Of Game Theory Application of Bargaining theory to the problem of resource allocation and call admission in heterogeneous wireless network in our contributed works. 01.01.2022 . Slide 18 Game Theory A form of mathematics which attempts to predict behavior in any sort of "strategic" environment For more on GAME THEORY… Kindly watch the Movie solution concepts for It develops proveable „The Beautiful Mind“…. negotiating in situation of conflict of interests…. 01.01.2022 . Slide 19 Bargaining 01.01.2022 . Slide 20 Online Bargaining Bargaining turns out to be a daily activity in most areas……. 01.01.2022 . Slide 21 Bargaining Players will gain if they agree on a solution, otherwise they will go back to their status quo. Different solutions have been proposed for bargaining problems e.g. Nash Bargaining solution, Kalai-Smorodinsky (varient of Nash) 01.01.2022 . Slide 22 Bargaining Problem Bargaining Problem = (S, d) U2(x) Convex and Compact S d2 d1 U1(x) S = feasible set Which value is the correct solution in Feasibility set……………….. 01.01.2022 d = disagreement point . Slide 23 Axioms of Bargaining Solution 1. Pareto Optimal A solution is pareto optimal if it is not possible to find another solution that leads to a strictly superior advantage for all players simultaneously U2(x) S S2 S1 U1(x) 01.01.2022 . Slide 24 Axioms of Bargaining Solution 1. Pareto Optimal 2. Affine Transformation 3. Symmetry 4. Indedependence of Irrelevant Alternatives 4. Individual Monotonicity 01.01.2022 . Slide 25 Bankruptcy General Bankruptcy 1. Our Bankruptcy Equ. n b 0 E ci 0 rpa (q) where C = (c1,…,cn) E = Requested Bandwidth wa w (q) i 1 n x 2. i 1 i C = Pre-defined offered bandwidth by Acc. Tech Question……………. E n * a x r i p (q) How should the resource be allocated………?? (where x = allocation) i 1 (where x* = allocation) Given that 0 E ci (rpa (q), B a (q)) (E, C) 01.01.2022 . Slide 26 Resource Bankruptcy as Bargaining To define bargaining problem associated with bankruptcy problem, we define a convex and compact feasibility set for resource allocation problem. S (rpa (q), B a (q)) {x a n | x a B a (a), x wa rpa (q)} wW The disagreement point in our problem formuation is influenced by cooperation among a a different access S ((rp (qto),one B operator. (q)),0) technologies belonging Disagreement Point = 0 01.01.2022 . Slide 27 Bargaining Problem 0-associated Bargaining problem [Proof ommitted] Proportional Distribution Rule F P r (( rpa ( q ), B a ( q )) (q) w(W W a ) a ( q ) r p (q) w(W W 01.01.2022 a a A ) . Slide 28 Bandwidth Request b a (q) b a (q) WLAN b a (q) Allocation w.r.t Pre-defined offered bandwidth…?? Any access technology getting into congestion will not be able to offer predefined offered bandwidth…. b a (q) a bba (q(q) ) b a (q) So we define the term „Offered BW“ WiMAX UMTS a3 WiMAX UMTS a4 a1 Offered BW Pre-defined Offered BW a2 WLAN 01.01.2022 . Slide 29 Offered Bandwidth Offered BW = Tuned by congestion factor UMTS y = lw C w lw ,therefore w (q) = bw (q)y WLAN Offered BW = Scaled by Loadsharing w (q) WiMAX 01.01.2022 where . = cw Ca = b k (q )(1 y ) kW a Slide 30 Offered Bandwidth So the allocation is……. X wa rp (q) b w (q) where b is proportionate factor 01.01.2022 . Slide 31 Putting things together… X wa rp (q) b w (q) B (q) = Wa if a b WiMAX UMTS rp (q ) a3 0 a4 rp (q ) otherwise Ca a1 a2 WLAN w (q) = 01.01.2022 bw (q)y if w W a cw Ca if w W a c wa if Wa 0 lw Cw lw y =e b (q)(1 y ) = k kW a . Slide 32 CAC, Mobility Algorithms X wa rp (q) = X ao ac r (q) = r p (q ) w (q) r p (q ) w(W W a ) 0 r p (q ) p Otherwise If rpac (q) w (W W 0 a ) Otherwise cwao (t s) cwao (t ) X ao rp (q) 01.01.2022 (q ) . Slide 33 Simulation Scenario - Area a1 is considered here. -Calls for different applications generated using poisson distribution with mean 7 -Call holding time: infinite 01.01.2022 . Slide 34 Mobility simulation -Simulated for Mobility between areas 01.01.2022 . Slide 35 Comparison -Our approach compared against the different approaches e.g. Best Fit, Worst Fit etc. [comparison paper – D. Mariz, I. Cananea, D. Sadok, and G. Fodor, “Simulative analysis of access selection algorithms for multi-access networks,”] 01.01.2022 . Slide 36 Thanks 01.01.2022 . Slide 37