This repository is a curated public research-code release for a memristor-based spiking neural network accelerator designed around a bio-inspired interception task. It includes selected MATLAB workflow scripts, Verilog-A model files, sanitized Spectre workflow templates, public figures, and documentation associated with the related papers.
Repository URL:
This project studies a bio-inspired interception task in which a predator-like agent uses motion-related information to predict and respond to a moving target. The task provides a compact benchmark for evaluating event-driven neural computation, closed-loop trajectory behavior, and hardware-aware implementation of spiking neural networks.
Rather than treating the SNN only as a software model, this work explores how the network can be mapped toward memristor/RRAM-based synaptic hardware and neuron-level VLSI circuits. The goal is to connect task-level behavior, SNN computation, device/circuit modeling, and system-level evaluation in a single research workflow.
The workflow combines software-level SNN modeling with hardware-oriented mapping and circuit simulation:
- Formulate the bio-inspired interception task using predator/prey position and motion variables.
- Use an SNN model to process task-related inputs and generate prediction or control outputs.
- Map trained or selected weights into hardware-oriented representations.
- Represent synaptic computation using memristor/RRAM-inspired crossbar structures.
- Use Verilog-A model files and Spectre workflow templates to support circuit-level evaluation.
- Analyze trajectory behavior, waveform-level results, and variation effects through MATLAB-based scripts.
The full circuit-level workflow depends on licensed Cadence/Spectre tools and a compatible 130 nm PDK/ODK environment. These resources are not included in this public repository.
This repository sits at the intersection of spiking neural networks, memristor-based computing, neuromorphic circuits, and VLSI implementation. It is intended as a curated public release of selected code, model files, workflow templates, figures, and citation metadata associated with the related papers.
Please cite the related papers if you use this repository, its workflow templates, figures, or ideas:
- Qianhou Qu, Sheng Lu, Sungyong Jung, Qilian Liang, and Chenyun Pan, "Compact and Energy-Efficient Memristive Spiking Neuromorphic Accelerator for Bio-inspired Interception Tasks," arXiv:2605.31141, 2026.
- Qianhou Qu, Sheng Lu, Liuting Shang, Jaihan Utailawon, Sungyong Jung, Qilian Liang, and Chenyun Pan, "Memristor-Based Spiking Neural Network Accelerator for Bio-inspired Interception Task," arXiv:2605.31299, 2026.
Machine-readable citation metadata is provided in CITATION.cff.
src/matlab/
MATLAB scripts used for SNN, crossbar, neuron, mapping, waveform-processing,
and result-analysis workflows.
src/veriloga/
Verilog-A model files used in the research workflow. These files are provided
for reference and simulation-template purposes.
src/data_examples/
Small public CSV/data examples that are safe to include.
spectre_templates/
Sanitized Spectre netlist and run-script templates. These require an external
Cadence/Spectre installation and a properly configured 130 nm PDK/ODK environment.
figures/paper_figures/
Manually curated public figures associated with the project. Private notes,
local paths, and unrelated project figures are excluded.
docs/
Documentation explaining the project flow, external dependencies, Spectre workflow,
figure descriptions, and citation information.
audit/
Source-to-public file map, figure inventory, release-curation notes, and public
repository audit records.
Start with:
- docs/overview.md for the high-level project scope.
- docs/project_flow.md for how the MATLAB, Verilog-A, and Spectre workflow pieces relate.
- docs/spectre_simulation_workflow.md for external circuit-simulation requirements.
- docs/figure_description.md for the included public figure inventory.
Internal release-curation and privacy-audit records are stored under audit/.
The MATLAB and Spectre-related files reflect the original research workflow, but some scripts expect generated simulator outputs or local tool configuration that are not part of this public repository.
Full circuit-level simulation requires external resources that are not included:
- Cadence Spectre
- a valid Cadence license
- a properly configured 130 nm PDK/ODK environment
- local model-library paths
- generated netlists and simulator output directories
The public templates use placeholders such as $SPECTRE_BIN, $PDK_ROOT, $SPECTRE_MODEL_DIR, and $CADENCE_HOME.
- Selected MATLAB scripts for software/circuit workflow analysis.
- Verilog-A device/model files used by the research flow.
- Small public CSV examples.
- Sanitized Spectre templates with placeholders instead of private paths.
- Manually curated public figures associated with the project.
- Documentation and audit files for public release preparation.
- Full paper PDF.
- Full PowerPoint slide deck.
- PDK/ODK technology files.
- Cadence installation files, license files, or environment files.
- Raw Spectre outputs, logs,
.printfiles, or simulator run directories. - MATLAB binary artifacts such as
.matand.fig. - Private paths, server paths, personal emails, or local setup files.
This repository is released under the MIT License. See LICENSE for details.
This repository is a curated public code/documentation release associated with the related research papers. It is not a complete end-to-end circuit reproduction package. Proprietary simulator files, PDK/ODK files, and licensed Cadence/Spectre resources must be supplied by users through their own authorized environment.