Swarm performance indicators and algorithms for logistics use cases
: Robot swarm controllers for automated storage and retrieval, and metrics to describe swarm performance

Student thesis: Doctoral ThesisDoctor of Philosophy (PhD)

Abstract

Robot swarms for logistics scenarios have been proposed many times but not successfully implemented with entirely distributed systems. They are hypothesised to be useful for use cases where minimal set up, infrastructure, cost and computational load are required.
This translates to a robot swarm for storage and retrieval that can be used out-of-the-box. We first conducted a use case study to perform mutual shaping of the research into swarm robotics for logistics by engaging with potential users of the technology before it is developed. These provided useful insights which informed the design of the following swarm algorithms and confirmed that automated storage and retrieval would be useful beyond large, industrial warehouses. Using the results from this use case study, swarm algorithms were developed for an out-of-the-box, robot swarm for logistics in an unmapped space. These algorithms were based on random walk for searching the storage space, probabilistic sampling to reshuffle the items in storage and two methods for directed return without global information or central control. These two directed return algorithms are the Biased Heading Behaviour and Swarm Diffusion-Taxis. When tested in simulation, they both successfully created a taxis effect towards a delivery area with minimal set up, costs and no global information required. The Swarm Diffusion-Taxis algorithm was successful in robot taxis towards multiple, dynamic and changing areas of interest. Finally, to help with the implementation of this technology in industry, metrics were developed for swarm traits: Robustness; Fault Tolerance; Scalability; and Adaptability. These are known as Swarm Performance Indicators (SPIs). These SPIs were justified by a literature review into the qualitative use of these terms and traits which informed their quantitative definition, as is given in this work. Also given with the metrics is a testing demonstration on a logistics use case and a decision-making algorithm for robot swarms. This shows how the metrics can be used in practise. Overall, the use case study, the swarm algorithms for logistics and the Swarm Performance Indicators contribute to the future implementation of swarm robotics into real world applications.
Date of Award19 Mar 2024
Original languageEnglish
Awarding Institution
  • The University of Bristol
SupervisorSabine Hauert (Supervisor) & Mahesh Sooriyabandara (Supervisor)

Cite this

'