Uncertainty in action selection and sequencing.

The execution of a complex task, comprised of multiple sub-tasks, is a frequent realworld
challenge. Uncertainty in this process will be familiar to anybody that has had
to assemble IKEA furniture for the first time. Sensory and perceptual uncertainty
may be reduced through goal-driven gaze shifts to acquire task-relevant information
at the right time. The planning of larger scale movements (e.g. reaching) under
conditions of uncertainty is thought to involve competition between multiple motor
programmes. The gradual resolution of this competition is reflected in characteristic
reach-paths. As such, movements of both gaze and limbs may provide insight into
the agent's level of uncertainty at a particular point in time during task execution.
However, while gaze has been studied in naturalistic sequential tasks, most of the
work on uncertainty in motor planning involves simple, "single-shot" movements. The
challenge taken up here in this project is to understand and model uncertainty at
different levels in the hierarchy of decision making, from goal and sub-goal planning
down to basic movements. We will introduce different forms of uncertainty in the
context of complex, sequential tasks and use continuous measurement paradigms to
identify uncertainty and its resolution from motor behaviour. Beyond basic theoretical
and empirical interest, this work has the potential to inform the fields of robotics and
human-computer interaction, as well as to supplement novel delivery methods for
training in technical and/or complex tasks.