Logo image
Back
An FPGA implementation of spiking neural network wavefront path planning
Thesis   Open access

An FPGA implementation of spiking neural network wavefront path planning

Thomas L. O'Keeffe
Master of Science (M.S.), Drexel University
Jun 2026
DOI:
https://doi.org/10.17918/00011391
pdf
OKeeffe_Thomas_20261,009.81 kBDownloadView
PDF Open Access Open Access (License Unspecified)

Abstract

Field-programmable gate arrays Robot path planning Spiking neural networks Wavefront propagation
Autonomous robots operating in unstructured environments require navigation systems that can adapt to changing terrain and internal states in real-time. Traditional path-planning algorithms, such as A* or Dijkstra's, often struggle with the computational overhead of frequent replanning on sequential, von Neumann architectures. This thesis presents SHAPE (Spiking Hardware-Accelerated Planning Engine), a Field-Programmable Gate Array (FPGA) implementation of the biologically-inspired navigation system proposed by Krichmar et al. SHAPE utilizes a Spiking Neural Network (SNN) based on the concept of cognitive maps, whereby environmental traversal costs are represented by axonal delays rather than traditional synaptic weights. Updating these delays through experience allows the system to react to dynamic environments without the need for global replanning. The hardware design exploits the inherent parallelism and local communication of SNNs to reduce the bottlenecks associated with data movement in standard processors. This work details the architecture of SHAPE, the trade-offs involved in its FPGA implementation, and the methodology used for its evaluation. Benchmarks show that SHAPE is 16x faster on average than A*, which suggests that this hardware-accelerated approach provides a viable path for real-time robotic navigation in complex settings.

Metrics

1 Record Views

Details

Logo image