Amnon Meisels
...
on our way to Changu Narayan...
my
Kids (from Back...)
and Full faces
am@cs.bgu.ac.il
Research
interestsThe general area of distributed constraints processing has been at the center of my research for the last 25years. The natural focal point of my group over the years was that of representing and solving Multi-agent Optimization Problems Distributed Constraint Optimization Problems (DCOPs) have served for two decades as a crystal clear laboratory for multi-agent search algorithms, due to the unique clarity of the problem definition - Vriables; Domains; (well-defined) Constraints. Our group's work produced the best performing concurrent search algorithms; concurrent measures of performance; and concurrent heuristics.
Having
produced the best performing distributed search algorithms on DCOPs, we have
moved our focus in the last ten years to Asymmetric DCOPs (ADCOPs). ADCOPs have
constraints among agents such that the gains of the agents from a given
constraint between them are not equal. In other words, the constraints matrices
become Bi-matrices. Consequently, the general constraints are equivalent to
normal form games among pairs of agents.
This lead to
focusing our research in two fields:
· Games
on Networks (GoNs) (ADCOPs JAIR 2013), where the two-agent
games are general bi-matrices
·
Distributed (multi-agent) search
algorithms on some specific GoNs:
o
Solving
Boolean Games by using incentives
The natural
next step was to consider an important family of games on networks that have
been studied extensively in both economics and sociology - Public Goods Games
(PGGs) and to designing multi-gent (distributed) search algorithms for Solving PGGs by using Incentives .
The major step forward was taken in 2018,
when the idea of adding side payments among neighboring agents on the network
was first added to search algorithms on ADCOP search by Selfish Agents and
became the
Iterative Nash Efficiency Enhancement (INEA) algorithm presented and
published in 2018.
In the same two years my group has also
worked on the Public Goods Game (PGG) which is a well researched example of a
Game on Networks. PGGs have played an important role in economics and in social
sciences for the last two decades. Our research on finding efficient solutions
for PGGs produced two major results. One, that PGGs are potential games (for
K=1). The other (and more important), is a search algorithm for PGGs that uses
side payments among agents and guarantees convergence to stable
PGG solutions with higher efficiency.