Lightcap/pudu-robot-operation-logs-bau-capstone-2026

PUDU Robot Operation Logs for Predictive Maintenance

This repository publishes the raw robot-operation database snapshot used by the Bahçeşehir University Faculty of Engineering and Natural Sciences capstone project Fault Prediction and Diagnostics from Robot Operation Logs. The project was completed during the 2025-2026 Spring semester as an interdisciplinary Computer Engineering and Industrial Engineering capstone.

The dataset supports predictive maintenance, fault detection, diagnostics, robot fleet monitoring, KPI reporting, and reproducible evaluation of LSTM and Random Forest models trained on PUDU robot operational logs.

DOI: 10.57967/hf/8635

What Is Included

Source table	Dataset file	Rows	Columns
model_training.training_artifacts	data/model_training_training_artifacts.parquet	3	7
model_training.training_runs	data/model_training_training_runs.parquet	1	20
model_training.training_source_tables	data/model_training_training_source_tables.parquet	5	5
public.robot_logs_error	data/public_robot_logs_error.parquet	147,488	23
public.robot_logs_error_test	data/public_robot_logs_error_test.parquet	22,124	23
public.robot_logs_error_training	data/public_robot_logs_error_training.parquet	103,241	23
public.robot_logs_error_validation	data/public_robot_logs_error_validation.parquet	22,123	23
public.robot_logs_info	data/public_robot_logs_info.parquet	3,893	18

The central log table, public.robot_logs_error, contains 147,488 raw error/event records from 46 robots, covering 2025-05-03 to 2026-02-26. The training, validation, and test split tables contain the partitioned model-development snapshot used by the LSTM training pipeline.

Project Team

Name	Department	Email	Project role
Faruk Alpay	Computer Engineering	faruk.alpay@bahcesehir.edu.tr	Dashboard/backend development, API integration, data product publication
Nazım Alp Batu Kardaş	Computer Engineering	alpbatu.kardas@bahcesehir.edu.tr	Dashboard/frontend implementation and application integration
Buğra Kılıçtaş	Computer Engineering	bugra.kilictas@bahcesehir.edu.tr	Model training repository and LSTM/Random Forest model development
Bilge Ece Şentürk	Industrial Engineering	bilgeece.senturk@bahcesehir.edu.tr	KPI framework, operational analysis, reporting
Serra Uysal	Industrial Engineering	serra.uysal@bahcesehir.edu.tr	Maintenance process analysis, evaluation support, reporting
Ayça Yeralp	Industrial Engineering	ayca.yeralp@bahcesehir.edu.tr	Success criteria, failure rules, operational reporting

Advisor	Department	Email	Profile
Assoc. Prof. Cemal Okan Şakar	Computer Engineering	okan.sakar@eng.bau.edu.tr	Academic profile
Assoc. Prof. Adnan Çorum	Industrial Engineering	adnan.corum@eng.bau.edu.tr	Academic profile

Model Context

The model-development code is maintained at github.com/DrGb24/pudu_bot_model_training. The companion dashboard repository is github.com/farukalpay/pudu-robot-operation-dashboard. The registered LSTM model uses 10-timestep sequences with 8 engineered features:

error_count, task_hour_num, day_of_month, day_of_week, robot_id_length, software_version_length, product_code_type, hourly_error_rate.

Registered test-set performance: Accuracy: 98.94%, Recall: 95.76%, Precision: 85.30%, F1: 90.23%, AUC-ROC: 99.33%. The model artifacts captured with the dataset are stored under model_artifacts/.

Field Context

The following image was captured by Faruk Alpay in Ljubljana, Slovenia, at a McDonald's on 2026-04-02 and is included as informal field context for the service-robot platform studied in the capstone.

Data Files

• data/public_robot_logs_error.parquet: full raw error/event log table used by the dashboard.
• data/public_robot_logs_error_training.parquet: training split used by the LSTM pipeline.
• data/public_robot_logs_error_validation.parquet: validation split used for tuning and early stopping.
• data/public_robot_logs_error_test.parquet: held-out test split used for final evaluation.
• data/public_robot_logs_info.parquet: raw robot information/status log records.
• data/model_training_*.parquet: training run registry, source-table counts, and artifact metadata.
• metadata/database_manifest.json: exported table schema, row counts, ranges, and snapshot metadata.
• metadata/data_dictionary.csv: table-column dictionary extracted from PostgreSQL metadata.
• metadata/kpi_framework.json: KPI definitions derived from the project KPI template.
• model_artifacts/: trained LSTM and Random Forest artifacts from the capstone training repository.
• reference/HATA_KODLARI_ROBOT.xlsx: robot error-code reference workbook from the training project.

Dataset Viewer Configurations

The Hugging Face Dataset Viewer is explicitly configured because the repository contains several raw PostgreSQL tables with different schemas. Each table appears as a separate viewer configuration, while the model-development split tables are grouped under partitioned_error_logs.

Viewer config	Split(s)	File(s)
full_error_log	full	data/public_robot_logs_error.parquet
partitioned_error_logs	train, validation, test	data/public_robot_logs_error_training.parquet, data/public_robot_logs_error_validation.parquet, data/public_robot_logs_error_test.parquet
robot_logs_info	full	data/public_robot_logs_info.parquet
training_runs	full	data/model_training_training_runs.parquet
training_artifacts	full	data/model_training_training_artifacts.parquet
training_source_tables	full	data/model_training_training_source_tables.parquet

Loading The Data

import pandas as pd
from huggingface_hub import hf_hub_download

repo_id = "Lightcap/pudu-robot-operation-logs-bau-capstone-2026"
path = hf_hub_download(repo_id=repo_id, repo_type="dataset", filename="data/public_robot_logs_error.parquet")
robot_logs_error = pd.read_parquet(path)

print(robot_logs_error.shape)
print(robot_logs_error[["robot_id", "task_time", "error_type", "error_level"]].head())

from datasets import load_dataset

full_logs = load_dataset(
    "Lightcap/pudu-robot-operation-logs-bau-capstone-2026",
    "full_error_log",
    split="full",
)

train_split = load_dataset(
    "Lightcap/pudu-robot-operation-logs-bau-capstone-2026",
    "partitioned_error_logs",
    split="train",
)

Dashboard Integration

The companion dashboard can read this Hugging Face dataset directly, removing the need for a live PostgreSQL connection for normal presentation and review workflows. The original PostgreSQL database remains the source system, while this repository is the citable, versioned publication snapshot.

Citation

@dataset{alpay_2026_pudu_robot_operation_logs,
  title  = {PUDU Robot Operation Logs for Predictive Maintenance},
  author = {Alpay, Faruk and Kardaş, Nazım Alp Batu and Kılıçtaş, Buğra and Şentürk, Bilge Ece and Uysal, Serra and Yeralp, Ayça},
  year   = {2026},
  publisher = {Hugging Face},
  doi    = {10.57967/hf/8635},
  url    = {https://huggingface.co/datasets/Lightcap/pudu-robot-operation-logs-bau-capstone-2026},
  note   = {Bahçeşehir University Faculty of Engineering and Natural Sciences Capstone Project, 2025-2026 Spring semester}
}

License

Released under CC BY 4.0 for academic citation and reuse. Please cite the dataset and the associated training repository when using this snapshot.