Back to search

University of Thessaly (UTH)

System support for the fault tolerance, testing and orchestration of drone applications

Abstract

dc:description

The continuous technological advances in embedded platforms and control systems have made unmanned aerial vehicles (drones) on the one hand more autonomous in their operation, combining computing capabilities, advanced navigation systems and obstacle avoidance systems, and on the other hand more affordable, paving the way for their use in many application fields, such as surveillance, monitoring and cargo delivery. Although drones can now be programmatically controlled via high-level commands, their smooth integration in the existing cyber-physical infrastructure is hampered by several factors including inherent safety and privacy concerns and the lack of automation in the (complete) drone operation lifecycle. To address this issue, in this thesis we present system-level mechanisms for strengthening the robustness of drone applications and promoting a more structured, managed and automated application development and deployment approach, properly integrated with the cloud and edge computing paradigms. In the first part of the thesis, we focus on tolerating failures critical to the target mission in application scenarios where a centralized mission controller entity is responsible for executing the application logic by coordinating a team of drones via high-level commands. We propose an active replication scheme combined with checkpointing and logging of communication and system status information for tolerating fail-stop failures, where the controller simply stops its operation. We show that in case of deterministic application programs the overhead introduced by our solution occurs only in case of drone failures and depends on the size of the logs that need to be exchanged between the controller replicas. Our solution also supports non-deterministic applications through a semi-active replication approach, in which case the overhead is analogous to the size of the execution state produced during the non-deterministic operations. Next, we use active replication with an agreement protocol to address Byzantine failures, the most general type of failures, where the controller behaves arbitrarily due to malicious attacks or hardware and software errors. Our approach relies on synchronous communication with signed messages and requires N = 2f + 1 replicas to tolerate up to f failures. We show that the introduced overhead comes mainly from the agreement protocol and can be greatly reduced by using a faster network for the communication between the replicas of the coordinator (a realistic assumption). Also, while the cost for typical setups is relatively high, at the range of tens of milliseconds, it is not prohibitive since the mission controller is not involved in tight control loops of the drones themselves. In the second part of the thesis, we look at how to assist drone application developers and users in the application deployment, monitoring and testing processes while ensuring that certain safety and privacy restrictions are not violated. To address these needs in a cost-effective manner, we adopt the platform as a service (PaaS) paradigm of cloud computing and view drones as a resource that can be accessed through a shared infrastructure. Firstly, we present a holistic approach towards supporting a more reliable managed operation of single-component drone applications through a software platform that takes care of the automated deployment and controlled execution, coupled with corresponding simulation and digital twin support. Differently to other efforts, the monitoring mechanisms do not concern only the "where" drones are allowed to operate, but also take account of the "how" their onboard sensing and actuation equipment is used. We also demonstrate the provided functionality through representative case studies using both the simulation environment where we perform a variety of tests before deployment, and the digital twin setup where a virtual representation runs in parallel with the application in the real world, providing the ability to evaluate, verify and even predict the behavior of the system. In the following, we expand various dimensions of the concept to distributed, component-based applications that span over the entire system continuum, including drones, edge nodes and the cloud. We present an orchestration framework that enables the high-level description of application requirements through suitable specifications that accompany the component code, supports the drone mission-aware deployment of the components, promotes the efficient communication between interacting components by transparently exploiting wireless networking technologies for direct communication and ensures safety- and privacy-preserving operations through the policy-based access to mobility and sensor services. We demonstrate the provided functionality via both field tests and the simulation environment. Further, through extensive performance evaluation, we show that our implementation has an acceptable resource footprint making it suitable even for the constrained computing platforms found on drones and edge nodes.

Degree

thesis:*
Grantor dc:publisher
University of Thessaly (UTH)
Year dc:date
2022

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Grigoropoulos, Athanasios
  • Γρηγορόπουλος, Αθανάσιος

Subjects

dc:subject × 16

Rights

Language dc:language
eng

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/51512

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Grigoropoulos, Athanasios; Γρηγορόπουλος, Αθανάσιος. System support for the fault tolerance, testing and orchestration of drone applications. University of Thessaly (UTH), 2022. http://hdl.handle.net/10442/hedi/51512