A Dynamic Observation Strategy for Multi-agent Multi-armed Bandit Problem.

Overview

We define and analyze a multi-agent multi-armed bandit problem in which decision-making agents can observe the choices and rewards of their neighbors under a linear observation cost. Neighbors are defined by a network graph that encodes the inherent observation constraints of the system. We define a cost associated with observations such that at every instance an agent makes an observation it receives a constant observation regret. We design a sampling algorithm and an observation protocol for each agent to maximize its own expected cumulative reward through minimizing expected cumulative sampling regret and expected cumulative observation regret. For our proposed protocol, we prove that total cumulative regret is logarithmically bounded.

@inproceedings{madhushani2020dynnamic,
  title={A Dynamic Observation Strategy for Multi-agent Multi-armed Bandit Problem},
  author={Madhushani, Udari and Leonard, Naomi Ehrich},
  booktitle={2020 19th European Control Conference (ECC)},
  pages={1677-1682},
  year={2020},
  organization={IEEE}
}