T-B PHASE overview
Autonomous machines generally work well in isolation. However, interactions with challenging environments, with other autonomous systems, and with people as part of hybrid human-machine teams creates uncertainty and some surprising behaviours.
The T-B PHASE collaboration between Thales and the University of Bristol has sought to understand the real-world behaviour of hybrid autonomous systems and create more robust and reliable approaches to their development and operation.
This website shares the learning of the T-B PHASE research projects.
Introducing T-B PHASE
T-B PHASE (the Thales-Bristol Partnership on Hybrid Autonomous Systems Engineering) was a £4 million collaborative research project between Thales and the University of Bristol that ran from 2017 – 2023. It was co-funded by Thales and the Engineering and Physical Sciences Research Council (EPSRC).
What is autonomy?
Autonomy refers to the ability of machines and collections of machines to decide their own actions based on information they collect.
Autonomy is not simply either present or absent though. Rather, there are degrees of autonomy that recognise the level of independence given to machines and the required level of input of a human user or operator. In general, a system becomes more autonomous by reducing its need for human oversight, taking on more complex decisions itself, or handling a wider variety of conditions.
What is an autonomous system?
An autonomous system contains interacting machines that work in tandem where at least one of the machines has a degree of autonomy.
Interactions in systems in general often produce complex new behaviours, and several T-B PHASE projects studied these effects in autonomous systems.
What is a hybrid autonomous system?
Within T-B PHASE, a hybrid autonomous system referred to autonomous machines interacting with humans or a complex environment.
Humans could include supervisors, operators, co-workers, or bystanders.
A complex environment might include buildings or varied topologies that interfere with movement, communications, or sensing capabilities.
Again, these interactions can lead to unforeseen behaviours or different modes of behaviour that are complex and, on first sight, seem counterintuitive. For example, limited communications can result in more efficient multi-drone search and rescue operations.
T-B PHASE examined the effect of human involvement and environmental limitations on autonomous machine performance. This included projects such as investigating how humans behaved when dealing with machine autonomy and automated information.
Responding to real-world interactions
The interactions present in hybrid autonomous systems and the complex behaviours they lead to cannot be avoided. Instead, the systems must be designed and developed to ensure they will behave in acceptable ways.
This is extremely challenging because it is hard to predict what a given system will do. The wide scope for machines to make their own decisions, the different ways humans might interact with the system, and the environment the system operates in all create considerable uncertainty.
The engineering challenge this presents is to identify acceptable behaviour in a hybrid autonomous system without having to anticipate every possible outcome.
How did T-B PHASE tackle this?
T-B PHASE conducted basic research in three areas related to this challenge:
Architecting autonomy
The design of autonomous systems often focusses on the new capabilities being introduced but the operational context and interactions are just as important.
We have shown how autonomous systems can be designed and architected in ways that include real-world interactions from the outset.
Autonomous machines
Machine autonomy is most often achieved using algorithms to make decisions about the machine’s next actions.
We have investigated how algorithm design influences the behaviour of autonomous machines and systems, especially teams of cooperating machines.
Human factors
Humans are a vital part of a hybrid human-machine team but the human response to working with machines is often overlooked.
We examined how autonomous machines and humans interact, and how such hybrid systems should be designed to produce clear, timely, and appropriate decisions and actions.
Each T-B PHASE research strand used challenges, examples, and tools developed in conjunction with Thales. High-level findings were fed back to the company, and new methods and tools developed to embed the learning in Thales.
This website continues the work of communicating T-B PHASE outcomes to a wider community by describing the key outcomes of each research project.
The work to develop autonomy continues, but T-B PHASE has created understanding of how to design effective and efficient hybrid autonomous systems in real-world environments to help establish the future direction of the sector.