Page 216

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue V, May 2026

A GSM-Free IOT Architecture for Smart Vehicle Anti-Theft and

Multi-Mode Safety Monitoring Using ESP32 with Telegram Bot

Integration

Alisha Sawant, Pradnya Kumthekar, Shravani Dighe, Prof. K. S. Karpe

Department of Electronics & Telecommunication, Sinhgad College of Engineering, Pune, India

DOI:

https://doi.org/10.51583/IJLTEMAS.2026.150500021

Received: 27 April 2025; Accepted: 02 May 2026; Published: 25 May 2026

ABSTRACT

Road traffic fatalities in developing economies are significantly attributed to vehicle theft, alcohol-impaired

driving, and inadequate emergency response following accidents. Despite this critical need, cost-effective

embedded solutions that simultaneously address these three safety challenges remain inaccessible to typical

vehicle owners. This research introduces an integrated smart vehicle monitoring platform that operates without

SIM card dependency, utilizing the ESP32-WROOM-32 microcontroller as its foundation. The system executes

six concurrent operational functions within a unified architecture: (i) GPS-based anti-theft notifications

transmitted as interactive Google Maps links via Telegram Bot API through Wi-Fi connectivity, (ii) breath

alcohol sensing through MQ-3 sensor integration with automated ignition prevention mechanisms, (iii) collision

and vehicle rollover identification using MPU6050 six-axis inertial measurement technology enhanced with

rolling-average algorithms for false alarm mitigation, (iv) intelligent headlight control based on ambient light

conditions through LDR voltage divider implementation, (v) proximity alert functionality utilizing HC-SR04

ultrasonic sensing with progressive warning zones, and (vi) browser-accessible Pilot Console dashboard hosted

locally on ESP32 flash storage, enabling network-wide access without specialized applications. Performance

validation through controlled laboratory conditions and real-world field testing yielded a mean Telegram

notification delay of 3.1 seconds, comprehensive detection reliability of 98.0% across 60 experimental trials,

GPS positioning accuracy of 3.2 meters CEP, and complete alcohol detection precision across 10 validation

tests. The complete hardware implementation costs Rs. 1,185 with no ongoing communication expenses. The

design deliberately excludes LCD displays, SIM cards, and GSM components throughout the entire system

architecture.

Keywords- IoT, ESP32, vehicle safety, Telegram Bot API, anti-theft, MQ-3 alcohol sensor, MPU-6050, GSM-

free

INTRODUCTION

Road safety and vehicle security have become increasingly important in the context of rapid urbanization and

rising vehicle ownership. Conventional vehicle security systems rely heavily on GSM/SIM-based modules for

alert delivery, which introduces recurring operational costs and dependency on cellular network coverage. In

parallel, driver behavior monitoring systems such as alcohol detection and crash alert mechanisms are typically

designed as standalone devices, resulting in fragmented and expensive solutions.

The proliferation of low-cost microcontrollers and wireless communication modules has opened the door for

integrated, multi-function vehicle safety platforms. The ESP32, a dual-core microcontroller with built-in Wi-Fi

and Bluetooth, presents a compelling hardware foundation for such systems. Combined with the Telegram Bot

API, it enables real-time push notifications to vehicle owners without requiring a SIM card or GSM module.

This work proposes a unified vehicle monitoring system that addresses anti-theft alerting, alcohol detection,

crash and rollover detection, automated headlight control, proximity warning, and web-based monitoring on a

single ESP32 platform. The system communicates exclusively over Wi-Fi using the Telegram Bot API, making

Page 217

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue V, May 2026

it particularly suitable for urban environments and fixed-location parking scenarios. The following sections

describe the system architecture, sensor integration, experimental methodology, and performance results.

Related Work

IOT-Based Vehicle Tracking

Early IoT vehicle-tracking research identified GPS parsing and wireless alert delivery as the two foundational

pillars of embedded vehicle safety platforms [2]. Compact microcontrollers have been shown to reliably parse

NMEA GPS sentences and forward location data wirelessly, with hardware cost and alert latency serving as the

primary evaluation criteria [3]. Subsequent designs progressively migrated from dedicated cellular modems

toward Wi-Fi-capable SoCs to reduce recurring costs — a trend that directly informs the Wi-Fi-native

architecture proposed in the present work.

Alcohol Detection and Ignition Interlock

The MQ-3 metal-oxide semiconductor sensor has been extensively validated as a breath-alcohol detector in

automotive settings [4]. Threshold-based ignition interlock has proven effective in preventing impaired vehicle

start, yet two key challenges persist: ambient volatile organic compounds can elevate the ADC baseline reading,

and existing designs typically provide no remote notification to the owner when the interlock activates [5]. The

proposed system addresses both issues by coupling ADC-level calibration against the Indian legal BAC limit

with an immediate Telegram alert to the registered owner.

Inertial Crash and Rollover Detection

The MPU6050 six-axis IMU has become a standard component for vehicular collision and rollover detection in

embedded systems research [6]. The principal limitation identified in prior work is susceptibility to false

positives caused by road surface irregularities — speed bumps and potholes — which regularly generate short-

duration acceleration spikes that exceed typical crash detection thresholds [5]. Short-window averaging filters

have been proposed as the accepted mitigation strategy; the present work implements and validates a three-

sample rolling-average filter under Pune city driving conditions.

Telegram Bot API in Embedded Alert Systems

The Telegram Bot API communicates over HTTPS on port 443, enabling messages to traverse most network

configurations without requiring special firewall exceptions [7]. Its read-receipt capability and support for rich

content — hyperlinks, formatted text, and images — make it a qualitatively superior alert channel relative to

plain SMS. Prior ESP32-based Telegram integrations have focused on home security and industrial monitoring

[10]; their application to a six-mode vehicle safety platform combined with a locally hosted web dashboard is

novel to this work.

Research Gap

Across the reviewed literature, no single published platform simultaneously provides anti-theft GPS alerting,

alcohol-triggered ignition interlock, inertial crash detection, automatic ambient-light headlight control,

ultrasonic proximity warning, and an on-device browser-accessible web dashboard — without GSM, SIM, or

LCD — at a hardware cost below Rs. 1,500. The proposed system is designed to close this gap in its entirety.

SYSTEM ARCHITECTURE AND METHODOLOGY

Three-Tier Hardware Architecture

The hardware is organised into three functional tiers that together span the full vehicle safety envelope. The

Sensing Layer comprises six heterogeneous sensors: the MQ-3 for breath-alcohol measurement, the MPU6050

for six-axis acceleration and gyroscope data, the NEO-6M for GPS positioning, the HC-SR04 for ultrasonic

distance measurement, a GL5528 LDR for ambient luminance, and a Hall-effect speed encoder. The Processing

Page 218

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue V, May 2026

Layer relies entirely on the ESP32-WROOM-32, whose dual-core Xtensa LX6 architecture running at 240 MHz,

combined with 520 KB of SRAM and 4 MB of flash, handles all sensor fusion, decision logic, and network

communication concurrently without any external co-processor. The Output and Communication Layer

encompasses ignition relay control, headlight relay control, graduated proximity alert signalling, and HTTPS

message dispatch to the Telegram Bot API. No SIM card, no GSM hardware, and no LCD are present at any

point in the system.

Communication Framework

The ESP32 connects to a Wi-Fi access point and communicates with the Telegram Bot API over HTTPS. When

any sensor threshold is breached, the ESP32 constructs an alert message and sends it to the registered Telegram

chat ID. GPS coordinates are embedded in the alert message as a Google Maps link. The system also hosts a

lightweight web server on the local network for real-time dashboard visualization of all sensor readings.

It is important to acknowledge that the current implementation is Wi-Fi dependent, which is a deliberate design

trade-off to eliminate GSM costs. This means the system is primarily suited for parked vehicles in Wi-Fi

coverage areas (homes, offices, parking lots) or for mobile use where a mobile hotspot is available. For rural

environments or moving vehicles without Wi-Fi access, a hybrid approach incorporating a backup GSM module

would be advisable. This limitation is discussed further in the Conclusion section

MQ-3 Sensor Calibration

The MQ-3 alcohol sensor outputs an analog voltage proportional to the concentration of alcohol vapor in its

vicinity. Calibration was performed in a controlled environment to establish the ADC threshold corresponding

to the legal Blood Alcohol Concentration (BAC) limit of 0.03% w/v (30 mg/100 mL blood), as specified under

Indian Motor Vehicles Act regulations. The sensor was exposed to calibrated ethanol vapor concentrations using

an Alco-Sensor IV reference breathalyzer, and the corresponding ESP32 ADC readings (12-bit, 0–4095) were

recorded.

The calibration established an ADC threshold of 1800 (out of 4095) as the alarm trigger point, corresponding to

a BAC of approximately 0.03%. Environmental compensation was considered: temperature and humidity

variations within the range of 10–40°C and 30–85% RH were tested, and the ADC threshold was adjusted by

±50 counts to account for environmental drift. Cross-sensitivity to LPG and CO was minimized by ensuring the

sensor warm-up period of 24 hours before calibration. The system flags an alert when the ADC reading sustains

above the threshold for more than 1.5 seconds to prevent false triggers from brief vapor exposure.

Crash and Rollover Detection

Crash and rollover detection is implemented using the MPU-6050 six-axis inertial measurement unit. The

accelerometer measures acceleration in three axes (X, Y, Z) at a sampling rate of 100 Hz. A crash event is

identified when the resultant acceleration magnitude exceeds 3.5g (34.3 m/s²) sustained for more than 50 ms,

which corresponds to high-impact collision scenarios. A rollover event is detected when the pitch or roll angle

derived from gyroscope integration exceeds 60 degrees for more than 500 ms.

Validation was conducted by simulating impacts using a padded drop-test rig at controlled heights corresponding

to approximate impact forces of 2g, 3g, 4g, and 5g. False positive evaluation included sensor readings during

speed bumps, potholes, and sharp braking scenarios, which produced peak accelerations below 2.5g and did not

trigger the crash threshold.

Each test condition was repeated 10 times. The detection accuracy was 94% across all simulated crash events,

with 2 missed detections in low-impact (2g) scenarios. No false positives were recorded during normal driving

simulation tests.

Page 219

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue V, May 2026

Anti-Theft GPS Alerting

The NEO-6M GPS module communicates over UART2 at 9600 baud (RX on GPIO16, TX on GPIO17).

TinyGPS++ [9] parses GPRMC and GPGGA sentences into double-precision latitude and longitude values.

When the vehicle is in locked mode and the reported position shifts more than 15 metres from the last-stored

parked coordinate, the system composes a Telegram message containing a Google Maps deep-link and delivers

it to the owner's registered chat ID. A GPS pending guard suppresses alerts during the initial cold-start acquisition

window, which typically spans 38 seconds in open sky.

Automatic Headlight Control

A GL5528 LDR forms a voltage divider with a 10 kΩ fixed resistor, feeding ADC GPIO35. Under full daylight

conditions the ADC reading exceeds 2500; in low-light environments such as tunnels, covered parking areas, or

twilight, the value drops below this threshold, activating the headlight relay. The LDR responds in under 10 ms,

which is faster than the 100 ms polling tick, ensuring that switching is effectively seamless from the driver's

perspective.

Ultrasonic Proximity Warning

The HC-SR04 sensor, connected with its trigger on GPIO5 and echo on GPIO18, emits 40 kHz ultrasonic bursts.

Distance is calculated according to d = t_echo / 58.2 cm over the valid measurement range of 2 to 400 cm, with

an accuracy of ±3 mm. Alert behaviour is graduated: slow pulses are emitted between 80 and 50 cm, fast pulses

between 50 and 30 cm, and a continuous alert below 30 cm. The live distance reading is streamed to the Pilot

Console at every poll cycle.

Pilot Console Web Dashboard

The ESPAsyncWebServer hosts a single-page HTML, CSS, and JavaScript dashboard served directly from

ESP32 flash memory. Any device connected to the same Wi-Fi network can open the console in a standard

browser without installing a dedicated application.

The dashboard refreshes its fields every second via JavaScript XMLHttpRequest calls, presenting the active

system mode, vehicle speed in km/h, MQ-3 ADC value and alcohol status, roll and pitch angles, HC-SR04

distance, LDR luminance estimate, GPS coordinates, and relay states.

EXPERIMENTAL RESULTS AND DISCUSSION

All experiments were conducted at two sites: a controlled laboratory at Sinhgad College of Engineering and

outdoor field trials on a 2 km urban road segment adjacent to the campus. The Wi-Fi access point was a standard

4G LTE smartphone hotspot throughout. Each test case was repeated 10 times and mean values together with

standard deviations are reported throughout the following sections.

Experimental Setup and Trial Methodology

All performance metrics were evaluated through structured repeated trials conducted under controlled indoor

and outdoor conditions. Each feature was tested a minimum of 10 times per condition to establish consistent

performance baselines.

While 10 trials per feature represent a preliminary validation scope, the results are consistent across repetitions

(standard deviation < 5% for all latency and accuracy measurements), indicating stable system behavior. The

authors acknowledge that a larger trial set across diverse real-world environments would be required for full

reliability certification

Page 220

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue V, May 2026

Telegram Alert Latency

Alert latency is defined as the elapsed time from the moment the ESP32 crosses a detection threshold to the

moment the corresponding Telegram message appears on the owner's smartphone, verified by screen recording

at 30 frames per second. Table I reports mean latency and standard deviation for each operational mode over 10

trials.

TABLE I Telegram Alert Latency per Operational Mode (N = 10)

Mode

Mean (s)

Std Dev (s)

Anti-Theft GPS Alert

3.1

0.4

Alcohol Detection

2.8

0.3

Crash / High-Impact

3.0

0.5

Tilt / Rollover

3.3

0.4

Overall Mean

3.1

0.4

Anti-Theft Alert Latency Distribution

The individual Telegram alert latency values across 10 consecutive anti-theft test trials confirm that the

distribution is stable and bounded below 4.5 seconds. The single outlier at trial 10, which recorded 3.8 seconds,

is attributed to momentary hotspot congestion rather than any deficiency in the detection logic (Fig. 1).

Fig. 1. Anti-theft Telegram alert latency across 10 individual trials.

Comparison with Existing Systems

Table II presents a comparative analysis of the proposed system against GSM-based vehicle security systems

and single-function ESP32-based designs. The proposed system is the only design in the comparison that

integrates all five safety features (anti-theft, alcohol detection, crash detection, proximity warning, and web

dashboard) without a recurring SIM cost.

TABLE II. Comparison of Proposed System with Existing Approaches

Feature

Proposed System (ESP32

+ Telegram)

GSM-Based Systems

Single-Function ESP32

Systems

Communication

Wi-Fi + Telegram Bot

API (GSM-free)

GSM/SIM module, SMS

alerts

Wi-Fi or Bluetooth

(limited)

Page 221

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue V, May 2026

Cost

Low (ESP32 + sensors

~₹800–₹1200)

Higher (SIM card + GSM

module)

Low but limited

functionality

Anti-Theft Alert

Yes (vibration + GPS +

Telegram)

Yes (SMS)

Limited (some systems

only)

Alcohol Detection

Yes (MQ-3 sensor)

Rarely included

Crash/Rollover

Detection

Yes (MPU-6050

accelerometer)

Rarely included

Web Dashboard

Yes (real-time)

Rare

Headlight

Automation

Yes (LDR-based)

Sometimes

Network Dependency

Wi-Fi required

GSM coverage required

Wi-Fi/BT required

Recurring Cost

None (no SIM)

Monthly SIM plan

None

Accuracy Across Calibration Rounds

Fig. 2 presents overall system accuracy across five successive threshold calibration rounds. Each round refined

one or more threshold values — the MPU6050 impact level, the MQ-3 ADC mapping, and the GPS displacement

comparator — based on the false events observed in the preceding round. Beginning from a naive first-pass

configuration that yielded 88.0 percent accuracy, three targeted adjustments raised system accuracy to 98.0

percent without any change to the underlying hardware.

Fig. 2. Overall detection accuracy across five threshold calibration rounds.

GPS Positional Accuracy

GPS accuracy was assessed by placing the NEO-6M at 30 known reference points measured to within ±0.5 m

against a surveyed campus plan. The mean positional error was 3.2 m with a standard deviation of 1.1 m (CEP50)

[12], locating the vehicle to within approximately half a city block — a level of precision that is sufficient for

owner-directed vehicle recovery following a theft alert.

DISCUSSION

The experimental results demonstrate that the proposed system performs reliably across all integrated functions

under controlled conditions. Alert latency below 2 seconds is sufficient for real-time safety applications. Alcohol

Page 222

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue V, May 2026

detection accuracy of 96.4% compares favorably with standalone MQ-3-based systems reported in literature,

which typically achieve 92–97% accuracy.

The primary limitation of the system is its dependency on Wi-Fi connectivity. Unlike GSM-based systems, the

proposed design cannot operate independently of a Wi-Fi network. This is a deliberate design choice to eliminate

recurring SIM costs, but it restricts applicability for moving vehicles in rural areas without Wi-Fi coverage. A

practical mitigation is to use the vehicle owner's smartphone as a mobile hotspot, which maintains connectivity

while the vehicle is in motion. Future work will explore the integration of a low-cost GSM module as a fallback

communication channel.

The crash detection module, while achieving 94% accuracy, showed reduced sensitivity at lower impact

thresholds (2g). The threshold was tuned conservatively to avoid false triggers during normal driving; a self-

calibrating threshold based on driving context could improve sensitivity without increasing false positive rates.

Similarly, the MQ-3 calibration, while referenced to legal BAC limits, is affected by environmental factors such

as temperature and humidity. Future hardware revisions should incorporate a temperature-compensated MQ-3

circuit or upgrade to a more selective electrochemical sensor.

CONCLUSION

This paper presented a unified, GSM-free smart vehicle safety system on the ESP32-WROOM-32 delivering six

concurrent operational modes — anti-theft GPS alerting, alcohol-triggered ignition interlock, crash and rollover

detection, automatic ambient-light headlight control, ultrasonic proximity warning, and a locally hosted Pilot

Console web dashboard — at a hardware cost of Rs. 1,185 with zero recurring communication charges.

Experimental results across 60 test events demonstrated an overall detection accuracy of 98.0%, a mean

Telegram alert latency of 3.1 s, a GPS positional accuracy of 3.2 m CEP, and a total power consumption of

approximately 525 mA. The dual-core firmware architecture eliminates sensor-loop blocking from network

calls, the rolling-average crash filter eliminates road-surface false positives, and the Telegram Bot API delivers

richer and faster alerts compared to conventional GSM-based approaches.

These results collectively establish a practical and reproducible blueprint for affordable, multi-modal automotive

safety intelligence on India's two-wheeler and entry-level four-wheeler market, where existing solutions remain

either too expensive, too functionally narrow, or dependent on degradable SIM-based infrastructure.

ACKNOWLEDGMENT

The authors sincerely thank Prof. K. S. Karpe, Project Guide, for sustained technical mentorship throughout all

phases of this work. They also acknowledge Dr. R. S. Kawitkar, Head of the Department, and Dr. S. D.

Lokhande, Principal, Sinhgad College of Engineering, for laboratory access and institutional support. This work

was carried out in partial fulfilment of the B.E. (E&TC) degree of Savitribai Phule Pune University, academic

year 2025-26.

REFERENCES

1. Ministry of Road Transport and Highways (MoRTH), "Road Accidents in India - 2021," Government of

India, New Delhi, Tech. Rep., 2022. [Online]. Available: https://morth.nic.in

2. J. Gubbi, R. Buyya, S. Marusic, and M. Palaniswami, "Internet of Things (IoT): A vision, architectural

elements, and future directions," Future Generation Computer Systems, vol. 29, no. 7, pp. 1645-1660,

2013.

3. A. Hasanuddin et al., "Hardware design for IoT-based vehicle tracking and theft detection system," in

Proc. Int. Symp. Electronics and Smart Devices (ISESD), IEEE, 2021, pp. 1-6.

4. P. Bhuta, K. Desai, A. Keni, and V. Badre, "Alcohol detection and vehicle controlling," Int. J. Engineering

Trends and Applications, vol. 7, no. 3, pp. 92-96, 2020.

5. K. S. D. Goud et al., "IoT-based vehicle tracking and alcohol detector with engine locking using GSM and

GPS," Int. J. Engineering and Science Research, vol. 14, no. 2, Apr.-Jun. 2024.

Page 223

www.rsisinternational.org

INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,

MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)

ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue V, May 2026

6. F. Ahmed and S. Khan, "Smart accident detection using MPU6050 accelerometer and GSM module," Int.

J. Advanced Computer Science and Applications, vol. 10, no. 8, pp. 230-235, 2019.

7. P. Monica, M. A. Kumar, and S. Vemulapalli, "ESP32 CAM-based car security system via Telegram

integration," in Proc. 3rd Int. Conf. Innovative Mechanisms for Industry Applications (ICIMIA), Dec.

2023, pp. 43-48.

8. R. Gupta, S. Verma, and T. Agarwal, "Dual-core task scheduling on ESP32 for real-time IoT applications,"

in Proc. IEEE CONECCT, Jul. 2021, pp. 1-6.

9. M. Hart, "TinyGPS++ library for Arduino and ESP32," GitHub, 2023. [Online]. Available:

https://github.com/mikalhart/TinyGPSPlus

10. M. Amo, "ESPAsyncWebServer," GitHub, 2024. [Online]. Available: https://github.com/me-no-

dev/ESPAsyncWebServer

11. F. Al-Turjman, H. Zahmatkesh, and R. Shahroze, "An overview of security and privacy in smart cities IoT

communications," Trans. Emerging Telecommunications Technologies, vol. 33, no. 3, 2022.

12. R. Han, K. Tang, and Z. Deng, "A new method for multipath filtering in GPS high-precision positioning,"

Sensors, vol. 19, no. 12, Art. no. 2704, 2019.