Research Challenge 1: Cascading Failure

Lead: Professor Eddie Wilson

Outline: A cascading failure occurs when the impact of a local perturbation or fault that either arises naturally or as the result of a malicious attack is propagated rapidly across a system’s connected components, amplifying its extent and severity.

Challenge: Understanding cascading failure in a hybrid autonomous systems context; designing to minimise cascading failure (Robustness); monitoring, anticipating and recovering from cascading failure (Resilience); towards design principles that enable managing and guaranteeing the extent of cascading failure (Regulation).

Research Challenge 2: Life Course Autonomy

Lead: Professor Arthur Richards

Outline: A key driver in the development of an autonomous system is its ability to respond to perturbation. However, system design typically involves consideration of relatively acute perturbations (e.g., component failure, extreme environmental conditions) rather than understanding design decisions in the context of the entire lifetime of the system.

Challenge: Understanding the impact on hybrid autonomous systems of post-deployment changes in operational environmental, task demands, etc.; designing to minimise sensitivity to these life course changes (Robustness); monitoring, anticipating and recovering from related failure or degradation of performance (Resilience); towards design principles that enable managing and guaranteeing over the life course (Regulation).

Research Challenge 3: Decentralised Decision Making

Lead: Professor Jonathan Lawry

Outline: Decentralised decision making in autonomous systems can bring significant advantages of efficiency, scalability, flexibility and robustness by exploiting scarce or noisy data and limited communication in rapidly evolving environments. However, across autonomous systems use-cases designers must ensure that systems are robust to noise and highly fault tolerant and that the very convergence properties that make distributed algorithms so effective, do not, under the wrong circumstances, amplify and propagate errors so as to cause whole-system failure, i.e., cause cascading failure.

Challenge: Understanding the impact on hybrid autonomous systems of decentralised information and decision making; designing to minimise sensitivity to decentralisation challenges (Robustness); monitoring, anticipating and recovering from decentralisation issues (Resilience); towards design principle that allow managing and guaranteeing despite decentralisation (Regulation).

Research Challenge 4: Hybrid Systems

Lead: Professor Jan Noyes

Outline: The deployment of autonomous systems to operate in close collaboration with humans challenges us to systematically identify the appropriate limits, nature and level of autonomy. We must consider how humans will react to different kinds of autonomous behaviour and how effective and comfortable they will be in scenarios where critical operational decisions are being taken in real-time without the immediate opportunity to vet or override them. We must develop operational protocols to help ensure safe and effective interactions with humans sharing the same space as, e.g., autonomous vehicles.

Challenge: Understanding the impact on hybrid autonomous systems of their being embedded as part of mixed ecosystems comprising people and machines with varying levels of autonomy; designing to minimise sensitivity to these hybrid issues (Robustness); monitoring, anticipating and recovering from issues arising from these hybrid issues (Resilience); towards design principle that allow managing and guaranteeing in a hybrid context (Regulation).