Task Planning and Control Synthesis for Multi-Agent Systems
Date Issued
October 2024
Author(s)
Advisor
Abstract
One of the major areas of research interest in the field of robotics has been the operation of autonomous
multi-agent systems in order to perform complicated tasks in dynamical environments. Additionally,
the combination of task planning with supervisory control for autonomous multi-agent systems is very
challenging and it has been a topic of increasing interest in the last decade, since it can provide a robust
framework for autonomous multi-agent task planning and execution. While task planning considers the
determination of discrete sequences of actions that are appropriately executed ensuring that the system
reaches given objectives, the supervisory control considers the orchestration of the control actions across
multiple systems to achieve a given set of high-level specifications.
A key requirement for the automatic task planning is in determining the appropriate abstraction models.
A state-of-the-art problem is to determine the appropriate formalism to encode high-level specifications
of tasks and the consequent methodologies of converting those into low-level action descriptions. Applications
of formal methods to optimization problems provide abstractions for modeling the continuous
agents’ actions in an appropriate discrete domain. Task planning problems have a wide variety of applications,
such as robotics, logistics, healthcare, traffic control, autonomous vehicles etc..
This dissertation introduces a novel methodology for automatic task planning of (possibly) heterogeneous
multi-agent systems utilizing concepts from automata theory in an appropriate form for the proposed
methodology. The modeling of system’s capabilities, constraints and failure modes is presented. The
resulting solution is guaranteed to be complete and optimal, while a heuristic approach is proposed for
size reduction of the discrete domain of the system.
This dissertation includes a formal analysis leveraging the power and effectiveness of discrete abstractions
models in solving complex task planning problems.
This dissertation provides industrial applications leveraging the proposed methodology to expound the
modeling power of the proposed framework, while demonstrating its potential applications in providing
solutions for the industry.
This dissertation introduces a new supervisory control framework for heterogeneous multi-agent control
synthesis with reactive failure-mode based reconfiguration. Using concepts from automata theory in
an appropriate form for the proposed methodology, this framework synthesizes the appropriate control
actions for the multi-agent system to fulfill the given task specification with the capability of on-the-fly
optimal reconfiguration of the task plan in case of failure occurrence.
multi-agent systems in order to perform complicated tasks in dynamical environments. Additionally,
the combination of task planning with supervisory control for autonomous multi-agent systems is very
challenging and it has been a topic of increasing interest in the last decade, since it can provide a robust
framework for autonomous multi-agent task planning and execution. While task planning considers the
determination of discrete sequences of actions that are appropriately executed ensuring that the system
reaches given objectives, the supervisory control considers the orchestration of the control actions across
multiple systems to achieve a given set of high-level specifications.
A key requirement for the automatic task planning is in determining the appropriate abstraction models.
A state-of-the-art problem is to determine the appropriate formalism to encode high-level specifications
of tasks and the consequent methodologies of converting those into low-level action descriptions. Applications
of formal methods to optimization problems provide abstractions for modeling the continuous
agents’ actions in an appropriate discrete domain. Task planning problems have a wide variety of applications,
such as robotics, logistics, healthcare, traffic control, autonomous vehicles etc..
This dissertation introduces a novel methodology for automatic task planning of (possibly) heterogeneous
multi-agent systems utilizing concepts from automata theory in an appropriate form for the proposed
methodology. The modeling of system’s capabilities, constraints and failure modes is presented. The
resulting solution is guaranteed to be complete and optimal, while a heuristic approach is proposed for
size reduction of the discrete domain of the system.
This dissertation includes a formal analysis leveraging the power and effectiveness of discrete abstractions
models in solving complex task planning problems.
This dissertation provides industrial applications leveraging the proposed methodology to expound the
modeling power of the proposed framework, while demonstrating its potential applications in providing
solutions for the industry.
This dissertation introduces a new supervisory control framework for heterogeneous multi-agent control
synthesis with reactive failure-mode based reconfiguration. Using concepts from automata theory in
an appropriate form for the proposed methodology, this framework synthesizes the appropriate control
actions for the multi-agent system to fulfill the given task specification with the capability of on-the-fly
optimal reconfiguration of the task plan in case of failure occurrence.
File(s)![Thumbnail Image]()
Name
Anatoli_Tziola_PhD_2024.pdf
Size
6.15 MB
Format
Adobe PDF
Checksum (MD5)
9b7b8324cb26a304c2bc16bddf74982d