Visually guided motion through complex environments

Thesis

The ability of animals to swiftly move through cluttered environments has inspired biologists interested in understanding its underlying mechanisms and engineers interested in applying its underpinning principles. To avoid colliding with obstacles, animals first need to detect that an object may be in their way. A decision then needs to be made about avoiding contact with the object so that locomotor strategies can be adjusted accordingly. In this thesis, I break down the problem of how animals negotiate clutter into three distinct, but related questions: 1) What information is needed to avoid collisions, and how do animals perceive it?; 2) how do animals choose between alternative paths to take?; 3) how are such choices implemented? I study these questions in two model species that inhabit open spaces but regularly encounter clutter: the homing pigeon (Columba livia) and the Picasso triggerfish (Rhinecanthus aculeatus). I frame their clutter negotiation behaviour as a process of sequential gap selection, allowing me to decompose it into components that can be independently evaluated. In Chapter 2, I use a combined experimental and modelling approach to model pigeons' steering behaviour algorithmically in a two-alternative forced-choice task. The birds are tracked flying inside a large hall whose open exit was separated from the release point by a curtain creating two vertical gaps. I find that their flights are best modelled by delayed proportional navigation, the same steering controller that has been used to implement all basic geometries observed in pursuit behaviour. In Chapter 3, I use a high-speed motion capture system to track pigeons flying through an artificial forest of vertical poles. I find bird gap choices are best captured by an autoregressive model, including information about the exit location, the cartesian distance between the closest obstacles, and the apparent visual gaps between all the obstacles ahead. In Chapter 4, I use a similar experimental paradigm to investigate how fish negotiate clutter, a behaviour which has never been studied in fish before. I find evidence of fish choosing clearances most directly aligned with their target based on egocentric visual cues. Finally, I show how these components can be combined to simulate realistic animal trajectories through clutter. Overall, the results of my thesis provide a valuable new framework for looking at how animals move through complex environments.

Authors: Pérez-Campanero Antolín, N

• DOI: 10.5287/ora-o0zymaz1x
• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Keywords: obstacle avoidance; visual guidance; fish; bird flight; gap negotiation; animal behaviour; motion capture; vision
• Subjects: Cognition in animals; Birds; Animal locomotion; Tracking (Engineering); Visual learning