Improved Accuracy of Robot Localization Using 3-D LiDAR in a Hippocampus-Inspired Model

TL;DR

This work extends Boundary Vector Cells (BVCs) to three dimensions by introducing vertical angular sensitivity and integrating 3D LiDAR data, enabling more accurate robot localization in complex spaces. The model combines a 3D BVC layer with a Place Cell Network (PCN), using a product-of-Gaussians firing rate that depends on distance, horizontal angle, and elevation, and learns place fields via competitive learning. Across four simulated environments with increasing vertical complexity, the 3D BVC model yields more distinct, unimodal place fields and substantially reduced spatial aliasing compared to a 2D baseline, while preserving performance in near-planar cases. These findings suggest that incorporating vertical information significantly improves navigation and mapping in real-world 3D spaces and highlight directions for adaptive vertical-layer selection and multimodal sensor integration.

Abstract

Boundary Vector Cells (BVCs) are a class of neurons in the brains of vertebrates that encode environmental boundaries at specific distances and allocentric directions, playing a central role in forming place fields in the hippocampus. Most computational BVC models are restricted to two-dimensional (2D) environments, making them prone to spatial ambiguities in the presence of horizontal symmetries in the environment. To address this limitation, we incorporate vertical angular sensitivity into the BVC framework, thereby enabling robust boundary detection in three dimensions, and leading to significantly more accurate spatial localization in a biologically-inspired robot model. The proposed model processes LiDAR data to capture vertical contours, thereby disambiguating locations that would be indistinguishable under a purely 2D representation. Experimental results show that in environments with minimal vertical variation, the proposed 3D model matches the performance of a 2D baseline; yet, as 3D complexity increases, it yields substantially more distinct place fields and markedly reduces spatial aliasing. These findings show that adding a vertical dimension to BVC-based localization can significantly enhance navigation and mapping in real-world 3D spaces while retaining performance parity in simpler, near-planar scenarios.

Figures (9)

• Figure 1: Firing patterns of two place cells in an environment with cross-shaped boundaries. \ref{['fig:multimodal_cell']} shows the firing pattern of an aliased place cell with four distinct areas of activation. \ref{['fig:unimodal_cell']} shows the firing pattern of a unimodal place cell.
• Figure 2: System Architecture. The BVC layer processes LiDAR boundary distances and agent orientation, supplying excitatory and inhibitory input to the PCN. The PCN applies recurrent inhibition to maintain a well-distributed place representation.
• Figure 3: Illustrations of the four test environments with varying wall orientations, designed to evaluate the model's 3D spatial encoding capabilities. The ceiling is transparent in the visualizations to allow a clearer view of the environments.
• Figure 4: A sampling trajectory plot of the agent's random walk over 4 hours, demonstrating dense and uniform coverage across the environment.
• Figure 5: Percentage of place cells with $\mathrm{MI} > 0$. Across all models and environments, the proportion of active cells remains consistently high.