Bioinspired lateralisation to improve search in robot swarms

点击翻译按钮获取中文摘要

"Lateralisation, colloquially referred to as handedness, is common in animals and its evolution suggests that it may improve efficiency in robotics. In groups such as bird flocks, fish schools and robot swarms, diversity in lateralisation may be particularly beneficial: having individuals that are more sensitive to stimuli on their left and others that are more sensitive on their right may give the group as a whole a greater ability to collectively detect stimuli than groups of non-lateralised individuals. In this PhD project we will test this in shoals of fish, combining the expertise in lateralisation of Prof. Culum Brown (Macquarie University) with an approach developed by Prof. Christos Ioannou (University of Bristol) to study search and detection during fish shoal collective behaviour (MacGregor et al., 2020 Nature Communications). Using findings from and guiding further behavioural work, the student will test how lateralisation in swarming robots affects the ability of robot swarms to detect stimuli using simulations and a swarm robotics testbed with Dr. Sabine Hauert (University of Bristol).The project has the following objectives: In fish shoals, determine the relationship between the mix of individual types within the group (right lateralised, left lateralised and non-lateralised) and the efficiency of responding to ephemeral resources. In fish shoals, quantify the effect of variation in lateralisation on collective movement, particularly shoal cohesion.Simulate the effects of variation in lateralisation in a generic model of collective movement, including how the agents respond to ephemeral resources.Implement findings from the simulation in real robots, using a swarm robotics testbed.We predict that a mix of lateralised (to both right and left) and non-lateralised individuals will be most effective for group cohesion, but groups without non-lateralised individuals will be most effective at searching. This is because non-lateralised individuals may act as a 'glue' to keep groups together, and tend to be found at the centre of groups, but contribute less to detection of targets.The project is well within the EPSRC remit as it seeks bioinspired solutions to improving the capability of robot swarms, which have a wide range of potential real-world applications. In particular, swarming robots tasked with searching for targets may be more effective when the swarms are a mix of individuals lateralised to the right, lateralised to the left, and non-lateralised. This project falls within the EPSRC Robotics and Control engineering research areas."