Tyche Special Research Initiative (SRI) on trusted autonomy

Project description

Our research program comprises the following sub-programs and involves technologies from computer science, software engineering and artificial intelligence. We are researching and developing techniques in classical planning, scheduling, optimisation and reasoning about nested belief (epistemic reasoning).

(i) Survivability, Explainability and Distributed Adaptability in Autonomous Systems

Autonomous systems are being increasingly considered in a range of applications and industries. However, current systems typically operate in constrained and well-defined environments, and therefore lack the ability to deal with fundamental uncertainty. Defence Science and Technology (DST) Group is interested in exploring the application of fully autonomous systems to Defence, but in their current state autonomous systems would simply be unable to operate in the dynamicity and uncertainty that are inherent to Defence environments. The Planning & Logistics (P&L) and Advanced Vehicle Systems (AVS) disciplines of DST Group are seeking to investigate autonomous systems through the themes of survivability, explainability and distributed adaptability. P&L is interested in the application of autonomy in Defence logistics and combat service support (CSS). In particular, autonomous CSS systems must be capable of operating in unstructured and uncertain environments, and must remain operational in the face of failure (survivability). At the same time, the human operators must be able to understand the decisions made by the autonomous system (explainability). The outcomes of this work will contribute to Project Tyche, a DST Group strategic research initiative on trusted autonomous systems.

(ii) Automated Reasoning Under Irreducible Uncertainty

DST Group’s Trusted Autonomous Systems SRI (Project Tyche) has identified that the inability of current approaches to adequately handle unpredicted and fundamentally unpredictable future states of the environment is the central problem to be addressed in trying to develop operationally deployable autonomous systems worthy of the name. The goal of this project is to advance the capabilities of automated reasoning as one necessary component of meeting this goal of producing systems that are able to cope with uncertainty, within operationally acceptable limits.

Logics and systems for formal reasoning have been around for many years, and underpin many familiar enabling technologies, such as circuit design and verification, language compilers, automated proof systems, databases, automated planning and scheduling, and in automated control systems. Yet the contention here is that significant potential remains untapped, because the goals of most automated reasoning efforts have been symmetric – maximum efficiency, ensuring complete input/output behaviour, etc – which assume stationarity and regularity of the problem situation. The problem is to begin to extend the existing logic and formal reasoning scientific basis under the plasticity imperative, which refers to the required property of being able to deal socially with irreducible uncertainty.

The Department of Computing and Information Systems, University of Melbourne is a recognised leader in autonomous planning in dynamic environments, and has deep knowledge in the required topics of automated planning, multi-agent systems, reasoning about action and change, epistemic reasoning and nested beliefs.

(iii) Autonomous Team Tactics Discovery with Automated Planning

This sub-project is a multi-institute collaboration looking at the future of team tactical behaviour for autonomous systems initially focused on multi-agent simulation. The project has two major threads:

  1. Specification of Team Tactical Behaviour (led by RMIT University)
  2. Autonomous Discovery of Team Tactical Behaviour & Investigation into Plasticity in Planning (led by University of Melbourne)

The aim of this collaboration is to tackle a large problem of designing behaviour for autonomous systems from two different perspectives. First, we propose to address the problem from an Agent Oriented Software Engineering perspective to develop a methodology for specifying (and in the future verifying and validating) complex team behaviour. Second, we propose to tackle the problem of team tactical discovery from an automated planning perspective.

Researchers

  • Tansu Alpcan, Associate Professor, Electrical and Electronic Engineering
  • James Bailey, Professor, Computing and Information Systems
  • Michael Cantoni, Professor, Electrical and Electronic Engineering
  • Airlie Chapman, Mechanical Engineering
  • Vincent Crocher, Mechanical Engineering
  • Peter Dower, Associate Professor, Electrical and Electronic Engineering
  • Robin Evans, Professor, Electrical and Electronic Engineering
  • Farhad Farokhi, McKenzie Research Fellow, Electrical and Electronic Engineering
  • Nir Lipovetzky, Research Fellow, Computing and Information Systems
  • Jonathan Manton, Professor, Electrical and Electronic Engineering
  • Chris Manzie, Professor &¬†MIDAS Director, Electrical and Electronic Engineering
  • Iven Mareels, Professor, Electrical and Electronic Engineering
  • Timothy Miller, Senior Lecturer¬†(Assistant Professor) & MIDAS Deputy Director, Computing & Information Systems
  • Girish Nair, Professor, Electrical and Electronic Engineering
  • Dragan Nešić, Professor, Electrical and Electronic Engineering
  • Denny Oetomo, Associate Professor & MIDAS Deputy Director, Mechanical Engineering
  • Marimuthu Palaniswami, Professor, Electrical and Electronic Engineering
  • Adrian Pearce, Associate Professor, Computing & Information Systems
  • Ben Rubinstein, Senior Lecturer (Assistant Professor), Computing & Information Systems
  • Dongryeol Ryu, Senior Lecturer (Assistant Professor), Infrastructure Engineering
  • Len Sciacca, Enterprise Professor, Electrical and Electronic Engineering
  • Iman Shames, Senior Lecturer (Assistant Professor) & MIDAS Deputy Director, Electrical and Electronic Engineering
  • Rohan Shekhar, Research Fellow, Mechanical Engineering
  • Liz Sonenberg, Professor and Pro Vice-Chancellor (Research Collaboration), Computing & Information Systems
  • Ying Tan, Associate Professor, Electrical and Electronic Engineering
  • Erik Weyer, Professor, Electrical and Electronic Engineering
  • Stephan Winter, Professor, Infrastructure Engineering
  • Miquel Ramirez, Research Fellow, Computing and Information Systems
  • Michelle Blom, Research Fellow, Computing and Information Systems
  • Sergio Jiménez Celorrio, Research Fellow, Computing and Information Systems

Partners

Defence Science and Technology Group

Research programs

Network Dynamical Systems