SDG Logos
Department of Computer Science and Engineering
Zeroth Review
Vehicle Problem detection
Student Members
Someshwar T.E - 310622104145 Supervisor:
Vischit Satish - 310622104172 Mrs.A.Jeba Sheela
Santhosh Kumar Y- 310622104130 Asst.Prof
Date: 12-07-2025 Zeroth Review 1
Aim of the work
• To develop a low-cost, sensor-based module for
real-time bike problem detection
• To enable early detection of engine, brake, and
motion-related issues using IoT sensors
• To notify users via app or buzzer before mechanical
failure occurs
Date: 12-07-2025 First Review 2
Objectives
• To build a plug-and-play
diagnostic unit using ESP32 and
external sensors
• To process sensor data using
edge machine learning for
predictive maintenance
• To integrate mobile app alerts
and optional emergency (SOS)
notifications
Date: 12-07-2025 First Review 3
Scope of the work
• Can be deployed in regular and electric
two-wheelers, both new and old
• Useful for delivery fleets, personal bikes, and rural
areas without access to smart diagnostics
• Scalable for accident detection, fleet health
monitoring, and connected mobility solutions
Date: 12-07-2025 Zeroth Review 4
Existing system 1:Harley Davidson Tool
About:
Harley-Davidson bikes (2001+) use an ECU to monitor systems. Tools like Motorscan
connect via a port to read faults and show live data on an app or laptop.
Algorithms:
-J1850 (PWM) protocol - UDS (Unified Diagnostic Services)
- CAN Bus (HD-LAN) - SAE J1979 (Mode 3 & 4)
- PID (Parameter ID) polling
Advantages:
Direct access to ECU data
Reads and clears fault codes (DTCs)
Disadvantages:
Reactive, not predictive
Limited to ECU-equipped bikes
No external sensor integration
5
Existing system 2:Ather System(E-Vehicle)
About:
Ather scooters use 40+ sensors with edge + cloud AI to monitor performance.
Real-time tasks run locally, while the cloud handles predictive maintenance and alerts.
OTA updates add features like fall detection and theft alerts.
Algorithms:
- Kalman Filter
- Recursive Least Squares (RLS)
- Signal processing(IMU)
Advantages:
Predictive Maintenance
Smart Features with OTA Updates
Disadvantages:
Brand-Limited & Not Universal
High Repair Costs & Service Issues
6
Base paper
● Title: “Cycling into the Workshop: Predictive Maintenance for
Barcelona’s Bike-Sharing System”
● Author name: Marcelo Arenas et al.
● Journal/Year:Springer Nature – AI & Society, April 2024
● Techniques:
– Survival analysis models (DeepSurv, CSF, MTLR)
– Predictive ML on historical trip and fault data
– Focus on brake, chain, wheel failures
● Pros & Cons
– Accurate predictions for scheduling repairs in shared bikes
– Lacks real-time sensing; not applicable to personal, non-smart
bikes Zeroth Review 7
Proposed System
Smart Bike Problem Detector using IoT and Edge AI
A low-cost, sensor-based embedded system designed to detect and predict faults in
any bike in real time. It collects physical data using vibration and motion sensors,
processes it locally with a TinyML model on an ESP32, and alerts the user through a
mobile app. Optional cloud support allows for analytics and historical logs.
Modules Used:
• Sensors: Vibration Sensor, MPU6050 (Accelerometer + Gyro)
• Controller: ESP32 (with Bluetooth/Wi-Fi)
• ML Model: Anomaly detection using Edge Impulse / TensorFlow Lite
• Communication: Bluetooth to App, optional GSM/Wi-Fi for cloud
• Mobile App: Health status, fault alerts, crash SOS
• (Optional) Cloud (Firebase/Blynk) for logs & fleet insights
Date: 12-07-2025 Zeroth Review 8
Proposed System
Date: 12-07-2025 Zeroth Review 9
Domain/Technology
● Domain: Intelligent Vehicle Monitoring & Predictive Maintenance in
IoT-enabled Mobility Systems
● Technology:
○ Edge AI / TinyML — On-device machine learning
○ Internet of Things (IoT) — Sensor integration (vibration, gyro,
temperature)
○ Embedded Systems — ESP32 microcontroller for data processing
and control
○ Mobile Application Development — Built using Flutter or Android
Studio
○ Optional Cloud Integration
Firebase/Blynk for storing fault logs and accessing analytics
Zeroth Review 10
Novelty of Proposed System
● Works on any bike — Not limited to EVs or smart bikes
● Uses Edge ML (TinyML) — Offline Access
● Real-time anomaly detection — Combines vibration and motion
patterns to detect engine faults
● Compact, single-unit design
● Crash detection with SOS feature
● Mobile app integration
● Low-cost, DIY-friendly
Zeroth Review 11
References
1. Marcelo Arenas et al. “Cycling into the Workshop: Predictive Maintenance for
Barcelona’s Bike-Sharing System”. EPJ Data Science, Springer Nature, 2024.
[Scopus Indexed]
2. R. Pupala, J. Shukla, “Review Paper on Vehicle Diagnosis with Electronic
Control Unit”. IJISRT, Vol. 3, Issue 2, 2018.
3. Ayush Gupta et al. “A Comprehensive IoT-Based Bike Crash Detection and
Emergency Response System”. arXiv preprint, 2022.
4. A. Sharma, S. Roy. “Low-Cost Industrial IoT System for Wireless Monitoring of
Electric Motors”. Mobile Networks and Applications, Springer, 2022.
5. Official documentation of Edge Impulse and TensorFlow Lite for
Microcontrollers, 2023.
6. Datasheets of MPU6050, SW-420 Vibration Sensor, and ESP32 DevKit V1.
Date: 12-07-2025 Zeroth Review 12
Thank You
Date Zeroth Review 13
