A General Framework for Charger Scheduling Optimization Problems

TL;DR

This paper introduces a deep reinforcement learning framework that automatically learns optimal charger scheduling strategies for wireless sensor networks, outperforming existing solutions across various NP-hard problems.

Contribution

The paper proposes a novel deep reinforcement learning-based framework that generalizes charger scheduling optimization, reducing problem-specific design and improving solution quality.

Findings

The framework effectively learns diverse scheduling strategies from experience.

Algorithms outperform existing methods in simulations.

The approach simplifies complex algorithm design for NP-hard problems.

Abstract

This paper presents a general framework to tackle a diverse range of NP-hard charger scheduling problems, optimizing the trajectory of mobile chargers to prolong the life of Wireless Rechargeable Sensor Network (WRSN), a system consisting of sensors with rechargeable batteries and mobile chargers. Existing solutions to charger scheduling problems require problem-specific design and a trade-off between the solution quality and computing time. Instead, we observe that instances of the same type of charger scheduling problem are solved repeatedly with similar combinatorial structure but different data. We consider searching an optimal charger scheduling as a trial and error process, and the objective function of a charging optimization problem as reward, a scalar feedback signal for each search. We propose a deep reinforcement learning-based charger scheduling optimization framework. The biggest advantage of the framework is that a diverse range of domain-specific charger scheduling strategy can be learned automatically from previous experiences. A framework also simplifies the complexity of algorithm design for individual charger scheduling optimization problem. We pick three representative charger scheduling optimization problems, design algorithms based on the proposed deep reinforcement learning framework, implement them, and compare them with existing ones. Extensive simulation results show that our algorithms based on the proposed framework outperform all existing ones.