Insect-inspired robot control

Some ant species are amongst the animal kingdom’s champion visual navigators. Despite having small brains and low-resolution vision, these ants can learn visually guided routes many metres long through complex terrain. However, in direct contrast with most modern robotic methods, insects use route knowledge not mental maps to navigate between two locations. That is, insects learn procedural instructions for navigation: “What should I do here?” rather than “Where am I?”. This allows for simpler representations of the visual world with corresponding potential for computational efficient bio-mimetic solutions for autonomous robotics. In the AI group at Sussex, we have explored how this type of navigational strategy can be implemented using simple mechanistic models, ANNs and spiking neural networks inspired by the architecture of insect brains. Here we present an overview of some of our group’s recent work on all three types of model, including demonstrating how temporal information can improve the performance and reduce the algorithmic complexity of mechanistic models, how early visual processing can improve model performance and, finally, how a spiking model of the Lateral Accessory Lobe —an area of the insect brain where pre-motor inputs are integrated — can elegantly incorporate active exploration when agents lose their way.

Short Bio

James Knight received his BEng degree in Electronic Engineering from the University of Warwick in 2006. After working in industry for several years, in 2013, he received an MPhil in Advanced Computer Science from the University of Cambridge and, in 2016, a PhD in Computer Science from the University of Manchester. His doctoral work focussed on using the SpiNNaker neuromorphic supercomputer to simulate large-scale computational neuroscience models with synaptic plasticity.

Since 2017, James has worked at the University of Sussex, focussing on using GPU hardware to accelerate bio-mimetic controllers for robotics and AI firstly as a Research Fellow and now as a Research Software Engineering Fellow funded by the EPSRC.

Back to the Schedule