English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

The ongoing deep learning revolution has allowed computers to outclass humansin various games and perceive features imperceptible to humans duringclassification tasks. Current machine learning techniques have clearlydistinguished themselves in specialized tasks. However, we have yet to seerobots capable of performing multiple tasks at an expert level. Most work inthis field is focused on the development of more sophisticated learningalgorithms for a robot's controller given a largely static and presupposedrobotic design. By focusing on the development of robotic bodies, rather thanneural controllers, I have discovered that robots can be designed such thatthey overcome many of the current pitfalls encountered by neural controllers inmultitask settings. Through this discovery, I also present novel metrics toexplicitly measure the learning ability of a robotic design and its resistanceto common problems such as catastrophic interference.  Traditionally, the physical robot design requires human engineers to planevery aspect of the system, which is expensive and often relies on humanintuition. In contrast, within the field of evolutionary robotics, evolutionaryalgorithms are used to automatically create optimized designs, however, suchdesigns are often still limited in their ability to perform in a multitasksetting. The metrics created and presented here give a novel path to automateddesign that allow evolved robots to synergize with their controller to improvethe computational efficiency of their learning while overcoming catastrophicinterference.  Overall, this dissertation intimates the ability to automatically designrobots that are more general purpose than current robots and that can performvarious tasks while requiring less computation.

Exploring the effects of robotic design on learning and neural control