A Reinforcement Learning Approach for Re-allocating Drone Swarm Services

TL;DR

This paper introduces a reinforcement learning framework for optimizing drone swarm re-allocation in delivery services, balancing provider profit and service constraints.

Contribution

It presents a novel RL-based method for dynamic drone swarm allocation considering environmental constraints and multiple consumer requests.

Findings

RL approach improves profit maximization

Efficient scheduling reduces run-time

Framework adapts to delivery environment constraints

Abstract

We propose a novel framework for the re-allocation of drone swarms for delivery services known as Swarm-based Drone-as-a-Service (SDaaS). The re-allocation framework ensures maximum profit to drone swarm providers while meeting the time requirement of service consumers. The constraints in the delivery environment (e.g., limited recharging pads) are taken into consideration. We utilize reinforcement learning (RL) to select the best allocation and scheduling of drone swarms given a set of requests from multiple consumers. We conduct a set of experiments to evaluate and compare the efficiency of the proposed approach considering the provider's profit and run-time efficiency.