WhiskEye: A biomimetic model of multisensory spatial memory based on sensory reconstruction

Knowles, Thomas C; Stentiford, Rachael; Pearson, Martin J.

Abstract

We present WhiskEye, a visual tactile robot supporting a neurorobotic investigation of spatial memory as a multisensory reconstructive process. This article outlines the motivation for building WhiskEye; the technical details of the physical robot, and the publicly available simulated platform via the NeuroRobotics Platform (NRP) from the Human Brain Project; and the biomimetic control architecture. The multisensory reconstruction model of place recognition based on deep predictive coding network is presented and datasets collected from the NRP are used to train and test the network. We demonstrate that the joint latent representations inferred by this system are positively correlated to displacements in pose space suggesting it is an advantageous sensory processing front-end for our neuro-plausible model of spatial memory.

• Presentation Conference Type: Conference Paper (published)
• Conference Name: TAROS 2021
• Start Date: Sep 8, 2021
• End Date: Sep 10, 2021
• Acceptance Date: Jul 1, 2021
• Online Publication Date: Oct 31, 2021
• Publication Date: Oct 31, 2021
• Deposit Date: Dec 8, 2021
• Publicly Available Date: Nov 1, 2022
• Publisher: Springer
• Volume: 13054
• Pages: 408-418
• Series Title: Lecture Notes in Computer Science
• Book Title: TAROS 2021: Towards Autonomous Robotic Systems
• Chapter Number: 43
• ISBN: 9783030891763
• DOI: https://doi.org/10.1007/978-3-030-89177-0_43
• Public URL: https://uwe-repository.worktribe.com/output/7931147