PUB - Publikationen an der Universität Bielefeld

# General Summary Spatial tasks, like recognising places, planning and optimising routes or estimating travel durations, permeate our daily lives and are navigational challenges shared not only by humans but most mobile animals. Spatial cognition research, the field of study associated with investigating these tasks, covers a diverse set of approaches, from neuro-physiological work, to behavioural studies and advanced computer-driven modelling. Research in this field provides valuable insights into the way mobile animals, including humans, perceive, navigate, and interact with their environments, offering the potential to enhance our understanding of fundamental cognitive processes and apply them to improve various aspects of our daily lives, from better urban design to the development of personalised technical assistance for individual navigators. In this dissertation, I present my contributions to the field of spatial cognition research, with a focus on the problem of returning to previously visited locations, also called *homing*. # Chapter Overview In the introductory chapter, I summarise a number of central concepts of the field and give a brief overview of the different ways animal navigators gather spatial information and use that information for navigation. In chapter 2, I introduce the ASV model, a vector-based image encoding algorithm, and outline how it can be used both for navigation and the analysis of image symmetry. I show that the ASV method's usefulness for navigation is closely linked to the local visual makeup of a location and its surroundings. In this context, I introduce the ”surroundedness” of locations as a possible indicator of their relevance for navigation. Specifically, I show that the average distance and spatial distribution of nearby objects might be an important feature that could be used as a predictor of how easily a location can be navigated back to. In the second part of the chapter, I show that the ASV can also be used to predict hovering locations of the marmalade hoverfly *Episyrphus balteatus*, highlighting the manifold applications of quantitative modelling approaches. Over the last years, research in spatial cognition has greatly benefited from the advent of virtual reality (VR) technology, allowing experimenters to design and present ever more diverse and interactive experiences. In chapter 3, I introduce the Virtual Navigation Toolbox [VNT](https://gitlab.ub.uni-bielefeld.de/virtual_navigation_tools), a collection of tools for the development of VR-driven spatial navigation experiments using the popular Unity game engine. The VNT prioritises modularity in its design, facilitating convenient implementation of diverse experimental paradigms. The VNT was also used to realise the experiment presented in chapter 4. In chapter 4, I investigate how human navigators combine information derived from their path integration system with different numbers of landmark cues in a series of virtual forest environments. I show that human navigators are best able to return to a goal location placed in between a small number of "landmark" objects and argue that the ability to successfully navigate in more cluttered environments depends, among other factors, on the ability to recognise the correct among multiple similar clearings in the forest surrounding the goal location. Therefore, successful navigation in clutter may depend on a navigator's ability to encode and recognise key "surrounded" locations, like the clearing marking the goal in this chapter or the more reachable locations surrounded by objects in chapter 2. I close out my dissertation with a brief summary of the main results of the previous chapters and discuss the implications of the presented results in the context of the ongoing debate on the different forms of spatial knowledge discussed in the field.