Model and Algorithm for Dynamic Multi-Objective Distributed Optimization

Abstract

Abstract. Many problems in multi-agent systems can be represented as a Distributed Constraint Optimization Problem (DCOP) where the goal is to find the best assignment to variables in order to minimize the cost. More complex problems including several criteria can be represented as a Multi-Objective Distributed Constraint Optimization Problem (MO-DCOP) where the goal is to optimize several criteria at the same time. However, many problems are subject to changes over time and need to be represented as dynamic problems. In this paper, we formalize the Dynamic Multi-Objective Distributed Constraint Optimization Problem (DMO-DCOP) and introduce the first algorithm called DMOBB to handle changes in the number of objectives.

Applicability of Multi-Agent Systems and Constrained Reasoning for Sensor-Based Distributed Scenarios: A Systematic Mapping Study on Dynamic DCOPs

“… Other scenarios were Dynamic DCOPs are modelled include collective autonomous vehicles for collision avoidance [ 66 ], human resources reorganisation such as scheduling elective surgery for emergencies in hospitals [ 67 ], distributed controlling of IoT devices and sensors for minimisation of energy consumption in smart homes [ 47 ], and dynamic social simulation through law enforcement problems and market-based mechanisms for dynamic task allocation to agents [ 68 ]. Proposals without scenarios [ 24 , 25 , 39 , 40 , 42 , 43 , 44 , 53 , 54 , 69 , 70 , 71 , 72 , 73 , 74 , 75 ] are the majority of the studies found. These studies refer to possible applications but are merely introductory examples without the context of a specific scenario.…”

“…Proposals without scenarios [ 24 , 25 , 39 , 40 , 42 , 43 , 44 , 53 , 54 , 69 , 70 , 71 , 72 , 73 , 74 , 75 ] are the majority of the studies found. These studies refer to possible applications but are merely introductory examples without the context of a specific scenario.…”

“…Machine learning agents on Dynamic DCOPs based on MDPs and Q-learning algorithms were also approached by [ 55 , 63 , 65 ]. Multi-Objective [ 69 , 71 , 72 , 75 ]: Okimoto et al modelled the Dynamic Multi-Objective DCOP (DMO-DCOP) as a sequence of static MO-DCOPs [ 69 , 71 ]. As explained before, Multi-Objective Optimisation consists of solving DCOP problems with more than one objective function that must be optimised simultaneously, with the possible existence of trade-offs among objectives.…”

“…These extensions are either inherited from other DCOP families or modelled as new functions and parameters to represent certain features described in the previous question: Multi-Objective Dynamic DCOP with algorithms such as DPR, DMOBB, and SSBDO [ 69 , 71 , 72 ]. Clement et al formalised first the DMO-DCOP problem as a sequence of a set of MO-DCOPs in the same fashion [ 71 ]. In their work, dynamics as changes in successive DCOPs were connected to the changes in the number of objective functions, whilst the number of agents, domains, and costs did not change.…”

“… Venue types: Workshops [ 21 , 42 , 69 , 74 ], Proceedings [ 17 , 18 , 22 , 24 , 25 , 39 , 40 , 43 , 44 , 46 , 47 , 53 , 54 , 56 , 57 , 59 , 60 , 61 , 62 , 63 , 64 , 65 , 66 , 67 , 70 , 71 , 73 , 75 ], and Journals [ 23 , 41 , 45 , 55 , 58 , 68 , 72 ]. …”