SmartStride: Revolutionising Treadmill Workouts

with IoT and Machine Learning

*Yashvi Agrawal, Hirak Desai, Khushi Dungrani, Diya Hirani, Sofia Francis

Abstract— This paper presents SmartStride, a novel smart transforming them from monotonous routines into
treadmill system integrating Internet of Things technology and personalised journeys towards peak performance.
machine learning to personalise the user's running experience. SmartStride stands at the forefront of this exciting
The system utilises an ESP32 microcontroller housed in a
movement. We envision a future where treadmills become
custom 3D-printed case attached to the treadmill. Infrared
intelligent companions, not just exercise equipment. By
sensors track the user's movement, and the collected data is
transmitted to a mobile application designed using MIT App seamlessly integrating Internet of Things (IoT) technology
Inventor. The app leverages user-specific parameters like and machine learning algorithms, SmartStride empowers
weight, height, and fitness goals to generate personalised run users to unlock their full potential, one personalised stride at
graphs and suggests workout plans via a machine-learning a time.
model. SmartStride addresses the growing demand for
intelligent fitness equipment by offering a user-centric This paper delves into the innovative features of
approach to treadmill workouts. SmartStride, exploring how it addresses the limitations of
traditional treadmills. We will examine the system
Keywords— Smart Stride, Internet of Things technology, machine
architecture, the role of machine learning in generating
learning, user-specific parameters, personalised run graphs,
fitness goals personalised workout plans, and the real-time data analysis
that fuels user progress. Through a detailed discussion of the
I. INTRODUCTION hardware components and software functionalities, we will
demonstrate how SmartStride bridges the gap between
The desire for a healthier lifestyle has fueled a constant human potential and technological capability. Finally, we
evolution in the fitness industry. From the rudimentary will explore the future directions for SmartStride, outlining
dumbbells of yesteryear to the high-tech exercise bikes of our vision for even more sophisticated user interaction and
today, technology has continually reshaped how we approach a truly holistic fitness experience.
our workouts. One area ripe for transformation is the
traditional treadmill. While offering a convenient and
efficient way to exercise, treadmills cannot often adapt to
individual needs. This one-size-fits-all approach can lead to II. LITERATURE REVIEW
frustration, plateaus in progress, and ultimately, a decline in
motivation. Numerous studies have delved into the realm of smart
treadmill innovations, aiming to revolutionise traditional
The inability of traditional treadmills to personalise the
treadmill training methods and enhance the overall running
workout experience stems from a lack of user data
experience. These endeavours encompass a wide array of
integration. These machines operate in isolation, offering
objectives and methodologies, each contributing valuable
pre-programmed settings that may not align with an
insights to the field of fitness technology.
individual's fitness goals, physical limitations, or running
form. This disconnect between the user and machine hinders
Researchers have explored the concept of automatic
progress and can even lead to injuries.
speed adaptation treadmills, seeking to address the
limitations of conventional treadmill training methods.
However, the winds of change are blowing through the
Through a comparative analysis of different approaches,
fitness landscape. The burgeoning field of smart
these studies aim to develop strategies for adapting treadmill
technologies is poised to revolutionise exercise routines,

Department of Computer Engineering, Mukesh Patel School of Technology Management and Engineering, NMIMS, Mumbai, India

yashvi.agrawal04@gmail.com, hirak214@gmail.com, khushi8dungrani@gmail.com, hiranidiya@gmail.com,

sofia.francis@nmims.edu
speed based on the user's desired pace, thereby optimising treadmill speed based on the user's desired pace. The
training effectiveness and user experience. methodology she involves is a review of existing literature
on automatic treadmill speed adaptation and a comparative
Another area of focus lies in validating algorithms for analysis of different approaches.
running analysis using wearable IoT technology. Studies
have aimed to assess the accuracy of inertial-based [2] Fraser Young, 2021, focuses on validating an inertial-
algorithms in quantifying ground contact time and swing based contact and swing time algorithm for running analysis
time during running, employing methodologies that involve using a foot-mounted IoT-enabled wearable device. The
recruiting large cohorts of participants for treadmill-based objective is to assess the accuracy of the algorithm in
testing while wearing IoT-enabled wearable devices. quantifying ground contact time and swing time during
running. The methodology includes recruiting a large cohort
Furthermore, research has been conducted on gait analysis of participants to run on a treadmill while wearing the
for shoe recommendation using wearable IoT technology. wearable device and comparing the algorithm's output with
By analysing gait patterns through inertial measurement unit manually labelled reference standard data.
(IMU) data, researchers aim to recommend suitable running
shoes based on pronation and foot-strike location. Data [3] Joao M.O. Passos, 2022, explores the use of wearable
collection from diverse cohorts of healthy individuals IoT technology for gait analysis and shoe recommendation.
running on treadmills with wearable devices is crucial to The objective is to develop an algorithm that analyses a
these studies. runner's gait patterns using inertial measurement unit (IMU)
data and recommends suitable running shoes based on
Comprehensive surveys have also been conducted to pronation and foot-strike location. He used a methodology
explore the landscape of wearable sensors and their that involved data collection from a large cohort of healthy
applications in gait analysis. These surveys aim to review adults and adolescents running on a treadmill while wearing
existing literature on wearable sensor technologies, data the wearable device.
processing techniques, and applications in gait analysis for
various purposes, ranging from rehabilitation to sports [4] Chad O’Brien, 2022, provides a comprehensive survey
performance and shoe recommendation. of wearable sensors and their applications in gait analysis.
The objective is to review existing literature on wearable
Moreover, comparative analyses have been undertaken to sensor technologies, data processing techniques, and
evaluate different types of self-paced treadmills for applications in gait analysis for various purposes, including
rehabilitation purposes. These studies seek to assess the rehabilitation, sports performance, and shoe
effectiveness of various control systems used in self-paced recommendation. The methodology involves a systematic
treadmills, such as position-based control, inertial force- review of relevant studies and a synthesis of key findings.
based control, and physiology-based control, by reviewing [5] Shaolong Liu, 2020, compares different types of self-
existing literature and analysing features and performance. paced treadmills for rehabilitation purposes. The objective is
to evaluate the effectiveness of various control systems used
Additionally, efforts have been made to develop wearable in self-paced treadmills, such as position-based control,
gait analysis devices for running injury prevention. inertial force-based control, gait parameter-based control,
Researchers aim to create devices capable of accurately and physiology-based control. The methodology includes a
analysing a runner's gait patterns and providing real-time review of existing literature on self-paced treadmills and a
feedback to prevent injuries. The methodologies employed comparative analysis of their features and performance.
include designing and prototyping wearable devices,
conducting validation studies to assess accuracy, and [6] Ravishankar Natarajan, 2022, introduces a wearable gait
evaluating effectiveness in injury prevention. analysis device for running injury prevention. The objective
is to develop a device that can accurately analyse a runner's
[1] Hee Jin Park, 2015, investigates the concept of automatic gait patterns and provide real-time feedback to prevent
speed adaptation treadmills, aiming to address the injuries. The methodology involves designing and
limitations of traditional treadmill training methods. The prototyping the wearable device, conducting validation
study aims to explore different methods of adapting
studies to assess its accuracy, and evaluating its
effectiveness in preventing running injuries.

III. PROPOSED SYSTEM

SmartStride introduces a ground-breaking approach to

treadmill fitness by seamlessly integrating various
technologies like IoT, Machine Learning, Cloud Computing
and Mobile Applications. The system, centred around an
ESP32 microcontroller, collects treadmill data, and
communicates via Bluetooth Low Energy. Additionally, IR
and gyroscope sensors enhance data accuracy. A user-
friendly Mobile App serves as the control interface,
facilitating interaction with backend systems and the IoT
device. Through machine learning, specifically LSTM Figure 1: Architecture Diagram
models, SmartStride analyses user data to craft personalised
workout plans stored in a cloud-based database. Heart rate 2. Mobile App
warning messages on the app screen during runs prioritise
user safety, ensuring a holistic and personalised fitness The Mobile App serves as the primary interface for users
experience. FastAPI ensures seamless data exchange, to interact with the SmartStride system. It provides controls
enhancing communication efficiency. Overall, SmartStride for starting, pausing, and stopping workouts, as well as
represents a significant leap in personalised fitness displaying real-time data such as speed, distance, and
equipment, delivering tailored workout experiences and calories burned. Additionally, the app facilitates
engaging users through its integrated ecosystem. communication with both the backend servers and the IoT
device, allowing users to access personalised workout plans
I. ARCHITECTURE
and track their progress over time.
The architecture of SmartStride is designed to
3. Machine Learning
integrate IoT technology, a mobile app, machine learning, a
database, and API calls to deliver a seamless and
personalized fitness Experience. [7] SmartStride leverages machine learning, specifically
LSTM (Long Short-Term Memory) models, to analyse the
data collected from users during their workouts. These
models process the data to understand patterns and trends in
the user's exercise habits, allowing them to generate
1. IoT (ESP32)
personalised workout plans tailored to individual fitness
The IoT aspect of SmartStride is facilitated by the ESP32 levels, preferences, and goals. This analysis is performed in
the cloud, enabling continuous improvement and adaptation
microcontroller, which acts as the central hub for collecting
of the workout plans based on user feedback and
live data on the treadmill. These sensors capture vital
performance metrics.
information such as running speed, distance, and inclination.
The ESP32 then communicates this data wirelessly over
Bluetooth Low Energy (BLE) to the mobile app, ensuring 4. Database
seamless connectivity with other system components, as
illustrated in Figure 1. The Database component of SmartStride stores a variety
of user-related information, including user profiles, workout
history, and system configurations. This data is stored
securely in the cloud, ensuring accessibility from anywhere
and facilitating seamless synchronisation across multiple
devices. By centralising this information, SmartStride
enables users to track their progress, review past workouts,
and adjust settings as needed to optimise their fitness
journey. II. DATA FLOW

5. API Calls The data flow in SmartStride outlines the path data
takes from collection at the IoT device through to storage
FastAPI serves as the backbone for facilitating data and retrieval in the cloud database.
exchange between different components of the SmartStride
system. It enables seamless communication between the 1. Data Collection at IoT Device (ESP32)
Mobile App, the cloud-based Database, and the Machine
Learning models, allowing for efficient retrieval and The ESP32 microcontroller, embedded in the accessory,
collects various data points from sensors such as speed,
processing of user data. This ensures a smooth and
distance, and inclination. These data points are processed and
responsive user experience, enabling SmartStride to deliver packaged into a structured format within the
personalised running patterns and insights in real time. microcontroller's memory.

2. Data Transmission via Bluetooth Low Energy (BLE)

The ESP32 microcontroller communicates wirelessly with

the Mobile App using Bluetooth Low Energy (BLE)
technology. The packaged data is transmitted as Bluetooth
packets every second. Contains relevant information about
the user's workout session.

3. Data Reception and Processing in the Mobile App:

The Mobile App, installed on the user's smartphone or

Figure 2: Circuit Diagram tablet, continuously listens for incoming Bluetooth packets
from the treadmill's IoT device. Upon receiving data, the
As shown in Figure 2 above, the IR Sensor Mobile App parses and interprets the information, extracting
(TCRT5000L) is a crucial component that tracks the speed of metrics such as speed, distance, inclination and heart rate. It
the treadmill by detecting a reflective marker placed on the then updates the user interface to display real-time workout
belt. This sensor enables precise measurement of the data, allowing the user to monitor their performance during
treadmill's velocity, providing accurate data for analysis and the exercise session.
feedback.
4. User Interaction and Feedback
The Gyroscope Sensor (MPU-6050) plays a vital
role in detecting the inclination of the treadmill. Measuring
The Mobile App also serves as the interface for user
orientation changes, it helps determine whether the treadmill
interaction, providing controls for starting, pausing, and
is flat or inclined, contributing to a more comprehensive
stopping the workout session. Users can customise their
understanding of the user's workout environment.
workout settings and preferences through the Mobile App,
The ESP32 Bluetooth Module serves as the central
such as setting target distances or adjusting incline levels.
processing unit for the IoT device. It collects data from the
Additionally, the Mobile App may provide feedback to the
IR and gyroscope sensors, processes it, and transmits it
user during the workout session, such as audio cues or visual
wirelessly to the accompanying mobile app. This module
notifications based on predefined thresholds or goals.
enables seamless communication between the treadmill and
the user's smartphone, facilitating real-time monitoring and
analysis of workout metrics. 5. Data Storage and Retrieval in the SQL Database

A reliable power source is essential to ensure the The Mobile App communicates with a remote SQL
uninterrupted operation of the IoT device. Using a portable database hosted in the cloud to store and retrieve user-related
power supply, such as a lithium polymer battery, guarantees data. At the end of each workout session, the Mobile App
continuous functionality of the sensors and Bluetooth uploads the collected workout data, including metrics and
module, allowing users to engage in their workouts without user settings, to the SQL database for storage. The SQL
any disruptions. database stores user profiles, workout history, and system
configurations, enabling users to track their progress over
time and maintain personalised settings across multiple significant weightage is assigned to data from recent runs. By
devices. prioritising recent run performances, SmartStride ensures
that workout plans evolve dynamically to align with the
6. Data Exchange via Python FastAPI user's progressing fitness journey. This approach not only
enhances the accuracy of the generated graphs but also
FastAPI facilitates communication between the Mobile facilitates the delivery of tailored workout experiences aimed
App, SQL database, and any other backend systems or at maximising user engagement and performance
services. It provides an efficient and secure mechanism for improvement. As a result, the graph generated on Day 1
transferring data between the Mobile App and the SQL (Figure 3.1) is very similar to the graph generated on Day 4
database, ensuring seamless integration and synchronisation (Figure 3.2), and likewise, the graph generated on Day 20
of user data. FastAPI handles API calls initiated by the (Figure 3.3) is similar to the graph generated on Day 26
Mobile App to retrieve workout plans, update user profiles, (Figure 3.4), highlighting how recent runs are given more
or fetch historical workout data from the SQL database. consideration.

III. DATASET IV. RESULTS AND ANALYSIS

The dataset component of SmartStride The SmartStride system was applied to a dataset
encompasses user profile data and previous run data to encompassing user profile data and information from
generate insightful graphs that guide workout plans. previous runs. The following analysis presents key findings
derived from this dataset.
1. User Profile Data 1. User Profile Analysis

The initial segment of the dataset comprises user profile Analysis of the initial graph generated from user profile
information, including age, height, and weight. This data data revealed that SmartStride effectively utilises
serves as the foundation for generating the first graph, which demographic variables such as age, height, and weight to
delineates initial workout parameters tailored to individual tailor personalised workout plans. This suggests that the
users. By analysing these demographic variables, system can adapt workout recommendations based on
SmartStride formulates a baseline fitness plan optimised for individual characteristics.
the user's physical attributes and health goals.
2. Analysis of Previous Run Data
2. User's Previous Run Data
Examination of data from users' previous runs indicated
After the establishment of the user profile, the dataset various trends and patterns. Metrics including distance
expands to incorporate data from the user's previous runs. covered, duration, average speed, and incline levels were
These data points encompass various metrics such as distance tracked. Notably, recent runs were given more weightage in
covered, duration, average speed, calories burned, heart rate, the analysis to reflect users' current fitness levels accurately.
and incline levels. Notably, the dataset prioritises recent runs,
placing greater emphasis on the user's current capabilities
and fitness level.

3. Graph Generation 3. Impact of Recent Runs on Graph Generation

Graphs are generated based on both the user's profile data Prioritising recent run data in the generation of subsequent
and their previous run data. The initial graph, derived from graphs ensured that workout plans accurately reflected users'
user profile information, sets the groundwork for evolving capabilities. By giving more weightage to recent
personalised workout plans as seen in Figure 3.1. runs, SmartStride effectively adapted workout
Subsequent graphs, however, focus predominantly on recent recommendations to align with users' current fitness levels.
run data. This emphasis on recent runs ensures that the
generated graphs reflect the user's current fitness status and
capabilities accurately.

4. Weightage Allocation

In the generation of subsequent graphs, particularly those

about workout plans and performance analysis, more
Figure 3.1: Graph generated by the model on Day 1 Figure 3.4: Graph generated by the model on Day 26
Table 1: Run data

Date Distance Time Run Volume

2024/03/15 1.30 km 10.13 mins 2612.79
2024/03/18 0.88 km 8.03 mins 1995.59
2024/04/03 2.75 km 11.8 mins 8448.00
2024/04/09 2.76 km 11.78 mins 9266.90

Incorporating the IR Sensor, Gyroscope Sensor, Bluetooth

Module, and power source into a custom 3D printed case
(Figure 4) made from PLA (Polylactic Acid) ensures a
Figure 3.2: Graph generated by the model on Day 4 compact and robust solution for SmartStride. This housing
not only protects the delicate electronic components but also
allows for easy attachment to the treadmill's side. By
integrating these elements into a unified enclosure,
SmartStride achieves a sleek and ergonomic design while
maintaining functionality and durability, providing users
with a seamless and hassle-free fitness experience.

4. Hardware results

Figure 3.3: Graph generated by the model on Day 20

Therefore, the evaluation of SmartStride's
personalised workout plans has demonstrated their efficacy
in meeting users' fitness objectives, as evidenced by the Figure 4: Design Prototype of the IoT Device
increasing run volume over time shown in Table 1. The V. LIMITATIONS
adaptive nature of these recommendations has contributed to
sustained user engagement and satisfaction. These findings While SmartStride presents a significant leap forward in
underscore the potential of SmartStride to revolutionise the
personalised fitness technology, there are limitations in the
fitness industry by providing tailored solutions that align
with individual needs and preferences. current prototype that future iterations aim to address. The
current system relies on infrared (IR) sensors to track user
movement, requiring a white marker on the treadmill belt, Diya Hirani: Supervision, project administration, funding
which can be inconvenient and visually distracting for some acquisition, writing—review and editing.
users. Future iterations will explore alternative sensor
technologies, such as embedded pressure sensors or camera- - Data Availability Statement: Not Applicable
based tracking systems, to eliminate this requirement and - Research Involving Human and /or Animals: Not
enhance user experience. Additionally, the mobile Applicable
- Informed Consent: Not Applicable
app controlling SmartStride and displaying workout data is
currently only compatible with Android devices, limiting
accessibility. Expanding compatibility to iOS devices is REFERENCES
crucial to ensure. [1] Hee Jin Park, Hyun Ju Lee, Tae Hun Kang, and Jeon IL
Moon, “Study on Automatic Speed Adaptation
VI. FUTURE WORK AND CONCLUSION Treadmills”, Division of loT & Robotics Convergence
Research, DGIST, 2015 15th International Conference
SmartStride ushers in a new era of personalised fitness on Control, Automation and Systems (ICCAS 2015)
technology, transforming the traditional treadmill into an
intelligent companion. By harnessing the power of machine [2] Fraser Young, Samuel Stuart, Rosie Morris, Craig
learning and seamless IoT integration, SmartStride Downs, Martin Coleman, Alan Godfrey, “Validation of
empowers users to achieve their fitness goals more an inertial-based contact and swing time algorithm for
effectively and enjoyably. The ability to generate running analysis from a foot mounted IoT enabled
wearable”, 2021 43rd Annual International Conference
personalised workout plans, track real-time data, and receive
of the IEEE Engineering in Medicine & Biology Society
tailored coaching fosters a more engaging and motivating (EMBC), DOI: 10.1109/EMBC46164.2021.9631046
exercise experience.
[3] Joao M.O. Passos, Sergio Ivan Lopes, Pedro M. Moreira,
As we look to the future, SmartStride's journey is far from Filipe M. Clemente, Pedro Bezerra, Rui Silva, Luis P.
over. One key area of focus is refining the machine learning Rodrigues, “An IoTenabled Portable System for Field-
model to provide even more sophisticated and data-driven based Endurance Run Testing”, DOI:
workout recommendations. Furthermore, compatibility with 10.1109/STAR53492.2022.9859636
popular fitness trackers like Fitbit and Apple Watch will
[4] Chad O’Brien, Cheol-Hong Min, “Classification of
unlock a treasure trove of user health data, enabling
Various Workout Motions Using Wearable Sensors”,
SmartStride to create personalised workout plans that DOI: 10.1109/AIIOT54504.2022.9817337
consider overall well-being. By relentlessly pursuing these
advancements, SmartStride aspires to become the gold [5] Shaolong Liu, Xingce Wang, Zhongke Wu, Ying He,
standard for smart treadmills, offering users a personalised “PhysioTreadmill: An Auto-Controlled Treadmill
journey towards peak performance. Featuring Physiological-Data-Driven Visual/Audio
Feedback”, 2020 International Conference on
- Competing Interests: Not Applicable Cyberworlds (CW), DOI:
- Funding Information: Not Applicable 10.1109/CW49994.2020.00042

- Author contribution: [6] Ravishankar Natarajan, Xinrui Yu, Jafar Saniie,

Yashvi Agrawal, Hirak Desai, Khushi Dungrani, Diya Hirani “Machine Learning Safety System for
Treadmill Users”, DOI:
10.1109/eIT53891.2022.9813874
The authors confirm their contribution to the paper as
follows:
[7] C.-M. Lee and C.-N. Ko, “Short-term load forecasting
using lifting scheme and ARIMA models,” Expert Syst.
Yashvi Agrawal: Conceptualization, methodology, data Appl., vol. 38, no. 5, pp. 5902–5911, May 2011.
curation, writing—original draft preparation, visualization.
Hirak Desai: Software development, validation, formal
analysis, investigation, writing—review and editing.
Khushi Dungrani: Resources, data curation, project
administration, writing—original draft preparation.