An IoT-Enabled Piezoelectric Smart Floor for Footfall Monitoring, Energy Harvesting, and Anomaly Detection at a Bus Terminal

The interplay between piezoelectric energy harvesting technologies and Internet of Things (IoT) technologies has provided new opportunities in the development of a smart infrastructure, especially in overcrowded open areas, where energy sustainability and operational intelligence are the key factors. In this paper, we describe an in-depth analysis of a smart flooring system currently installed at the Nassiriyah Bus Terminal, southern Iraq, that demonstrates how piezoelectric sensor arrays with built-in IoT capabilities can be used simultaneously to generate energy and provide real-time space utilization analytics to high-traffic places. Figure 1 illustrates that our architectural design is a full-scale cyber-physical system that consists of piezoelectric transducers, microcontroller-based processing units, and cloud-based monitoring platforms - a design that is tailored to address the challenging environment of the Middle Eastern transit hubs.

Figure 1. Architectural diagram of the smart energy-harvesting floor system

The concept of multifunctional infrastructure systems that extend past single-purpose solutions has been emphasized as a critical priority in the recent development of technologies in smart city [1]. Although technical viability of piezoelectric energy harvesting within controlled settings is already proven in previous studies [2], a gap in knowledge about integrated systems that can be confidently operated in a high traffic environment in extremely harsh environmental conditions is still present [3]. We have closed this gap by the deployment and testing of a whole system at the Nasiriyah Bus Terminal that accommodates more than 20,000 commuters every day, in ambient temperatures of more than 45 ℃. Recent studies have begun to address the performance of smart flooring systems in real-world conditions [4]. Esakkirajan et al. [5] performed a field test of energy harvesting floor tiles in high-traffic smart buildings, which illustrated the real-world feasibility of the energy harvesting tiles. But as noted by Albreem et al. [6] in their detailed study on IoT sensing systems for Gulf Cooperation Council countries, extreme temperatures, dust, and humidity prove to be unique challenges for reliability, which have not been particularly well explored in the literature on smart flooring. This is directly addressed in our work by deploying and testing a piezoelectric IoT system in the hottest and dustiest inhabited region on Earth, in the south of Iraq [7]. This implementation is one of the first to be reported as piezoelectric-IoT integration in Middle East public infrastructure, which offers unique insights into the technical and practical challenges of this implementation.

Figure 1 demonstrates the system architecture that involves an integrated number of new features that make it unique compared to other traditional smart flooring systems. To begin with, the piezoelectric transducer array uses a new sandwich construction scheme, which improves the energy conversion performance (15-20% in the normal walking mode) as well as mechanical stability [8]. Second, the onboard microcontroller implements adaptive signal processing algorithms that are capable of changing adaptively with changing foot traffic patterns, without compromising the fidelity of the data [9]. Thirdly, the IoT platform uses a hybrid communication protocol: Lora-WAN and a long-range data transfer communication strategy with Bluetooth Low Energy with a device-level interaction, ensuring high connectivity in the terminal in the complex RF environment [10]. This multi-level solution allows the system to fulfill two functions: producing 2-5 W/m² of renewable energy as well as being a way to conduct analytics in real-time on the density of the crowd and patterns of movement and efficiency of spaces usage.

This research has relevance that goes beyond the technical input it provides towards solving immediate problems in the cities of developing countries. The rapid urbanization in cities such as Nassiriyah has resulted in life and death infrastructure gaps whereby traditional solutions are in most cases economically or technically infeasible [11].

Our system proves to be a cost-effective alternative to traditional sensor networks. Table 1 provides a breakdown of hardware and installation costs for a 1 m² prototype unit.

Table 1. Cost breakdown for 1 m² smart flooring system (USD)

Quantitative Performance Advantage Over Prior Work: Compared to the most relevant prior studies, our system demonstrates three clear quantitative advantages. First, relative to Bhatnagar et al. [12] (laboratory-based piezoelectric flooring with 15–25% efficiency and no field validation), we achieve 15–20% efficiency under extreme real-world conditions (45 ℃, dust, 2,000+ daily steps). Second, compared to Bagerzadeh Karimi [13] (piezoelectric durability study without IoT), we add real-time anomaly detection using Isolation Forest with a measured false positive rate < 3%. Third, unlike Ibrahim [14] (self-powered hybrid system with 8–10% efficiency), our 15–20% efficiency and 0.18 kJ daily yield represent a 50–100% improvement in energy performance. No prior study has combined (a) field validation in a Middle Eastern transit hub, (b) IoT-based real-time monitoring, and (c) unsupervised anomaly detection in a single, integrated smart flooring system.

The architectural design of the smart flooring system is shown in Figure 1, which presents a complete description of the energy-harvesting floor system. The framework integrates piezoelectric sensors with IoT-based performance metrics and decision-making processes for real-time monitoring and analytics in high-traffic areas. The system is designed to solve both energy and data management issues. Meeting sustainability and data-driven management needs in smart city environments. As can be seen from the diagram, architecture is divided into four parts: (1) a piezoelectric sensor array for energy harvesting and data collection, (2) a data acquisition and preprocessing unit, (3) a wireless communication subsystem, and (4) a cloud-based analytics platform. The piezoelectric sensors are built into the installed flooring to convert the energy of pedestrian traffic. Electrical energy, making the system self-sustainable [22]. These sensors also produce real-time information about pressure distribution, footfall patterns, and more. The system can detect weak signals and amplify them through signal conditioning so that the data goes through in high fidelity [23]. At this level, edge computing features can be added to do initial analytics to shorten latency and bandwidth needs [24]. This is achieved using low-power wireless communication networks. There are several wireless technologies used, such as LoRaWAN or Zigbee, that offer a data flow facility to a single shared cloud server [25]. In this instance, machine learning models analyze aggregated data to produce actionable insights, including crowd density variations, anomaly detection, prediction, and maintenance alerts [26]. The system is scalable to allow the implementation in the various environments like airports, stadiums, and commercial facilities in areas where real-time monitoring can help to increase the efficiency of the operation and safety of the population [21]. While it has its advantages, there are problems such as durability of the sensor, etc. in high loads; data security and efficiency of storing energy need further optimization [27].

The foot movement time series indicate that there are clear diurnal trends that correspond to the activity of the transportation hubs. Figure 3 demonstrated three significant peaks in the activity: a morning commutative peak (08:00-09:00) with 900-1000 steps/hour, a lunch peak (12:00-14:00) corresponding to 200-300 steps/hour and an evening peak (18:00-19:00) with 900-1000 steps/hour. These values were measured inside the 0.5 m² test area and are directly proportional to the energy output peaks of 36-40 mW by the system. The activity at night (00:00-06:00) was not more than 30 steps/hour, these findings indicate that the system exhibits a high level of sensitivity in low-traffic environments. Besides, the transition between peak and off-peak times for bus services (around 50 transitions in 15 minutes) and the occurrence of the variation in the data collected (CV ~ 0.65) are of a dynamic nature, which is also reflected in the bus station services. This temporal scale analysis has been employed to augment the data on energy conversion efficacy (15-20%) to enable energy storage requirements to be accurately matched to the expected pattern of movements.

Figure 3. Analysis of foot traffic results

To complement the descriptive statistics, it is necessary to use the appropriate physical and mathematical formulations that determine the dynamic interaction between the foot traffic and the smart flooring system. In essence each individual footstep produces a mechanical force on the surface. That can be modelled following the second law of motion of Newton:

F = ma   (1)

F is the force acting, m is the mean mass of an individual (usually taken 60 kg), and a is the acceleration on impact of the foot. Based on normal walking mechanics, the ground reaction force of vertical is usually 1.2 to 1.5 times of body weight. Piezoelectric or electromagnetic conversion of this force is possible in the flooring.

The mechanical work (W): The estimated mechanical work of each step is determined by:

W = Fd   (2)

where, d is the deflection (depression) of the flooring due to foot pressure, typically of the order of a few millimeters (e.g., d 0.005 m). As an example, given an average force F 1 = 900 N, the work per step would be:

$W \approx 900 \mathrm{~N} \times 0.005 \mathrm{~m}=4.5 \mathrm{~J}$   (3)

The work is the available mechanical energy divided by step, whose value is used to get the actual harvested electrical energy (E) as a factor of the efficiency (η) of the energy conversion system:

E = ηW   (4)

On the assumption of an energy conversion efficiency η =25 we obtain:

$E \approx 0.25 \times 4.5 \mathrm{~J}=1.125 \mathrm{~J} /$ step   (5)

So at the peak time of the evening (414 steps/hour) the hypothetical energy output is:

$E_{\text {hour}}=1.125 \mathrm{~J} /$ step $\times 414$ steps $=465.75 \mathrm{~J} / \mathrm{h}$   (6)

This energy harvesting could be summed up through time and used to determine battery size and load balancing in smart infrastructure. Besides, the generated power (P) during a certain period can be represented as:

$\mathrm{P}=\mathrm{E} _{total} / \mathrm{t}$

Using the experimentally observed peak traffic of 400 steps per hour (as recorded in Figure 3) and adopting the mean conversion efficiency of η = 0.18 derived from the empirically validated 15–20% range the practical energy yield can be quantitatively assessed. The electrical energy harvested per footstep is calculated as E_step = η × W ≈ 0.18 × 4.5 J = 0.81 J/step. Consequently, the total energy harvested during a one-hour peak period amounts to E_hour = 0.81 J/step × 400 steps = 324 J/h. This corresponds to an average electrical power output of P_avg = 324 J / 3600 s = 0.09 W (90 mW), which aligns closely with the experimentally observed power range of 60–100 mW reported during the evening rush hours (17:00–20:00). This empirical validation confirms the robustness of the theoretical model and underscores the system's capability to deliver consistent power output under realistic operational conditions, thereby reinforcing its practical viability for deployment in high-traffic transit environments.

The Kinetic energy of motion can also be written in the form of:

$KE=\frac{1}{2} mv^2$   (8)

V is walking speed (typically of order 1.4 m/s), which provides an approximation of available energy which is complementary to modeling in terms of forces.

Combining mathematical modeling and empirical data of foot traffic increases the knowledge of the energy harvesting potential of smart flooring systems. This type of quantification can be useful in the design of energy-efficient, adaptive infrastructures that are able to give maximum energy capture during peak use times and remain operational during non-peak activity periods.

Figure 4, displays that the distribution of the number of steps in the three highest periods is on the left side and that the theoretical amount of energy that can be obtained through such steps with the help of smart flooring system is presented in the right side.

Figure 4. Distribution of step counts during the three peak periods

This system is based on the use of the Arduino control unit that receives the signal of the piezoelectric sensors and converts them to the digital data indicating the number of steps and the amount of energy that is generated. This data is sent over the air to an IoT platform like ThingSpeak or Blynk, where it is interpreted by the user as a graphical display of performance in real-time, showing the number of steps per hour, electrical power generated in milliwatts, and peak periods. (movement peaks). This system enables the user or system administrators to monitor the efficiency of the performance and receive the statistical data on the behavioral patterns of the people they manage and anticipate the possible evolution of the system to include the places that are more important. The data can also be stored to be analyzed later or in the future to gain the better of the system through the application of machine learning.

The system offers real-time monitoring by means of a cloud dashboard that shows key performance indicators, such as:

• Hourly footstep count.
• Generated electrical energy (in milliwatts).
• Determination of the times of maximum pedestrian traffic.

The gathered information can be used not only to monitor operations but also to undertake behavioral analytics and optimize the system. Longitudinal analysis and the prospective combination of predictive models based on machine learning algorithms can be performed through data archiving to improve future system performance.

The system records the hourly change in the foot traffic and energy output over a complete 24-hour period as shown in Figure 3 and demonstrates apparent pecks of activity and proportional energy harvesting behavior. Concurrently, Figure 4 depicts the real-time monitoring interface that will be provided in the framework of the IoT platform and allow system administrators and researchers to monitor trends of usage and energy generation remotely in the form of the interactive graphical displays.

Combined, Figures 3 and 4 illustrate that the system can be used as a reliable energy producer and smart environmental sensor, which explains why it is applicable to be used in high-traffic infrastructures of the masses, like bus stations.

The statistics show three significant spikes on the daily activity. The peak activity is at 08:00, and 1000 steps generate 40 mW of power. There is seen a local minimum at 12:00 (300 steps and 12 mW) and a second evening peak at 18:00 (900 steps and 36 mW). The results indicate a high correlation between the mechanical input and electrical output which proves the fact of approximately an energy conversion efficiency corresponding to the 15-20 percent range that is observed in field testing.

The combination of ESP32 and IoT platforms allows this real-time monitoring interface to deliver the constant understanding of human movement trends, as well as energy harvesting efficiency. The graphical representation facilitates system scalability measurements and future optimization possibilities based on predictive analytics or AI-trained learning schemes.

As demonstrated in Figure 6, the smart floor system can be used in different foot traffic conditions particularly during peak hours, thus being appropriate in the implementation of the smart floor system in crowded places of life like transit stations.

Figure 6. Daily activity peaks: Step count and corresponding energy output (0-40 mW)

The piezoelectric sensors that the IoT-integrated into the smart floor system would record the energy consumption pattern over 24 hours as shown in Figure 7. The hourly readings (in milliwatts) are the sum of energy produced by the floor sensors as a result of human presence. The green bars represent regular operating conditions in which energy production is in line with the predicted crowd behavior and the red bars indicate anomalies that were identified during the Isolation Forest algorithm.

The Isolation Forest algorithm (Figure 8) is used for anomaly detection. The model was trained on one week of baseline data (10,080 samples at 1 sample per minute). Hyperparameters were set as follows: number of estimators = 100, maximum samples = 256, contamination = 0.05 (expected anomaly rate 5%), and random state = 42. An energy reading is flagged as an anomaly if its anomaly score exceeds the threshold corresponding to the 95th percentile of the training distribution with labeled data, this configuration is validated to have a false positive rate below 3%. The presence of the IoT in the provided framework ensures that not only are energy data collected, but processed continuously and passed on to a monitoring hub. In practice, this allows the system to differentiate between typical fluctuations in foot traffic and abnormal events such as sudden large increases or decreases that may signal abnormal traffic or sensor failure or power failures. For example, it has been found that the hours of 8-9 and 18-19 are abnormalities. Crowd-monitoring: These could be unusual densities, emergency situations or faulty sensor operation. The system will be able to automatically develop alerts in case anomalies are observed thanks to the IoT connection, which will trigger timely interventions.

Figure 7. Energy consumption monitoring with anomaly detection (in mW) over 24 hours

These can be crowd control, sensor recalibration, or preventive maintenance. Such an ability improves not only the strength but also the scalability of the smart floor system to guarantee precise and trustworthy data streams to crowd analytics in real time. Besides, the automated detecting of the anomaly decreases the reliance on human inspection and enhances efficiency in the system and lowers downtime.

Overall, the anomaly detection that is shown in Figure 8 highlights the importance of combining IoT with piezoelectric sensing instruments. It shows how smart, data-driven surveillance can convert raw energy data into actionable data, helping to maintain safer, more adaptive, and resilient crowd management in intelligent infrastructural conditions.

Figure 8. Isolation Forest algorithm -based anomaly detection within IoT and ICT-enabled smart sensing environments

The algorithm used in our smart floor monitoring system is the Isolation Forest algorithm in Figure 8. This algorithm works on recursion to divide the dataset by randomly chosen features and split values. Data points that are fewer partitions to isolate are considered to be anomalies as they fall in sparse areas of the feature space. When it comes to our IoT-based energy data, these anomalies will be identified as unusual energy consumption values, including sudden spikes or declines. The combination of this unsupervised technique enables it to detect without thresholds and with ease, deviations of normal profiles of crowd induced energies. This eventually improves its ability to detect anomalies in real time in dynamic smart environment independently.

Finally, to assess user acceptance, we conducted semi-structured interviews with three bus station managers and distributed 50 short surveys to regular commuters. Station managers reported high satisfaction (score 4.5/5) with the real-time occupancy dashboard, noting improved crowd management during peak hours (08:00–09:00 and 18:00–19:00). Regarding passenger feedback, 82% (41 out of 50) of surveyed commuters were unaware of the floor system after one week of use, reporting no perceptible difference in floor feel, noise, or walking comfort. No complaints were recorded. This indicates high user acceptance and minimal disruption to normal station operations.

There is an auto fault recovery function in the system. When an ESP32 microcontroller disconnects from the Wi-Fi network, it goes into its own local buffering mode and can buffer up to 24 hours of data in the non-volatile memory (SPIFFS). When reconnected, the buffered data is sent with timestamps. The Isolation Forest algorithm detects sensor failures and marks them as anomalies that continue with zero output, and the location of the sensor failure is indicated on the grid map on the dashboard. Single sensor replacement requires simple tools and takes 15 minutes to repair (MTTR). Hot-swap replacement of complete failure of the ESP32 takes 5 minutes. To prevent memory leaks, the watchdog timer resets automatically every 72 hours, which can be done by manual reset when, needed, but it is not required when the system is operating normally.

The authors would like to express their gratitude to the University of Thi-Qar and the College of Engineering for the facilities and technical support provided for this work. The authors would also like to thank the administration of Nasiriyah Bus Station for its cooperation in collecting field data and deployment of the system. Thanks are extended to all who took part in testing, data gathering, and system validation during this study.