Intelligent system for detecting vehicle intrusion into restricted road areas

Abstract

Intelligent system for detecting vehicle incursions in restricted areas of roads, delimited by traffic cones on the roadway, consisting of two devices equipped with a two-channel radio system that creates a secure wireless communication network between cones and operators: (a) A compact, portable and autonomous intelligent traffic cone device, attached to the road cones as an "add-on" component, which integrates a sensor module combining an accelerometer and gyroscope and an AI module for the detection of intrusions and impacts; and, (b) an operator signal receiver device, to be worn by operators via a bracelet, with an appropriate interface for receiving alarms and controlling the system.This system detects the unusual movement of cones as a result of a vehicle entering the protected area, generating an immediate alert for workers and providing an effective safety solution for street and road maintenance.

Description

Intelligent system for detecting vehicle intrusion into restricted road areas The object of the present invention is an intelligent system for detecting vehicle incursions into restricted access areas of roads delimited by traffic cones on the roadway, implemented with ITS technologies of intelligent transport systems for road safety, in such a way that by means of motion sensors integrated into a device of specific design coupled to the cones and an AI application developed for this purpose, it is possible to detect the unusual displacement of the cones as a result of the intrusion of a vehicle into the protected area in order to generate an immediate alert to the workers. This intelligent safety system is based on two devices equipped with a two-channel radio that create a secure wireless communication network between traffic cones and operators in the application areas: (a) A compact, portable, and autonomous intelligent traffic cone device, attached to standard roadway cones as an "add-on" component, which integrates a combination accelerometer-gyroscope sensor module and an AI artificial intelligence module for the detection of intrusions and impacts; and (b) an operator signal receiver device, to be worn by workers in the construction area via a bracelet, with a user interface suitable for receiving alarms and controlling the system. A network configuration of this kind, between road marking cones and operators, offers an effective safety solution in street and highway maintenance works, by detecting and immediately alerting operators of critical situations in which a vehicle accidentally invades the restricted work zone, avoiding the false alarms generated by cone overturning detectors. SCOPE OF APPLICATION. - The technical field of application of this invention is that of traffic warning systems and devices regarding accidents and dangerous road conditions, and more specifically, that of intelligent detection systems for vehicle incursions into restricted areas. In parallel, the invention is also related to the related fields of testing and devices for measuring acceleration-deceleration in crashes, and wireless signaling and alarm systems. STATE OF THE ART. The concept of intelligent transport systems (ITS) is a set of technological solutions from telecommunications and computer science designed to improve the operation and safety of land transport, both for urban and rural roads, as well as for railways. With current technological advancements, there has been an evolution towards the design, development, and implementation of new applications that improve the efficiency, safety, and dissemination of information among the agents that integrate them, that is, between vehicles and / or infrastructure, through the exchange of data using various technologies and that can be applied to examples such as the dissemination of traffic conditions, information on the various road signs, or informational aids to improve the driving experience. Due to the special requirements of these types of applications, recent years have seen a rapid evolution in the emergence and development of new technologies that support them. This is more specifically known as vehicular communications, which can be further categorized into two subgroups: V2X, encompassing V2V (Vehicle-to-Vehicle) and V2I (Vehicle-to-Infrastructure) communications, and I2V (Infrastructure-to-Vehicle). In any case, the needs can vary depending on the type of application, ranging from low latency and high safety (braking systems or autonomous driving) to others with fewer restrictions (notification or communication of the status of points of interest). This leads to a situation where a wide array of technologies and solutions are available, making it difficult to obtain a specific, robust solution that meets the specific needs of market niches, such as, in this case, the road maintenance and operation sector. In road maintenance and operation activities, carried out in open traffic situations, there is a clear need to preserve the physical safety of the operators against incursions into their work area by vehicles traveling on the roads. The key difference between workplace accidents in road maintenance and other work environments lies in the fact that road workers, in addition to the common risks found in other areas of the construction and service sectors (injuries from tools, falls, injuries due to poor posture, etc.), are exposed to incidents caused by vehicles traveling on the roads. Specifically, this factor can lead to vehicles entering work zones closed to traffic, typically marked by traffic cones. The concern regarding the rate of workplace accidents in this field stems from the fact that these incidents are generally quite severe in terms of their impact on workers, often even resulting in death. Given the importance of this issue, the road maintenance sector faces a clear need and market demand for effective systems that enhance worker safety. Similarly, there are other areas where it is crucial to prevent various elements from exceeding the established boundaries of work zones, such as construction sites involving heavy machinery (dumpers, excavators, etc.), and industrial and logistical processes that utilize moving machinery, like pallet loaders. Currently, technological solutions exist for the intelligent detection of vehicle incursions into restricted road zones based on optical sensor systems or vehicle motion detection. However, these systems present several drawbacks, as they are affected by weather conditions and operate outdoors, constantly exposed to environmental changes. Additionally, optical sensor systems or motion detection systems are directly linked to the installation of sensors on semi-fixed elements that are not sufficiently versatile; in other words, in all cases, it is ultimately necessary to place traffic cones. Other solutions on the market rely on machine vision technology and image processing, achieving very good results and high accuracy. For example, a Spanish company's system consists of a device adaptable to a telescopic mast or surveillance vehicle. This device carries a high-definition recording camera connected to a communication station managed by video analytics software, which alerts workers to the entry of a vehicle into the work area through flashing lights on their vests and audible alarms. However, basing a safety system solely on this technology is insufficient due to its limitations in low-visibility conditions. In contrast, a system based on inertial sensors, such as the one in the present invention, guarantees vehicle detection through the cone network. Therefore, the proposed system consists of accelerometer sensors that selectively detect, via an AI application, the movement of road cones that constitute an intrusion, generating an immediate alert for workers. This system offers a real alternative to inefficient or limited tools, as it is unaffected by unforeseen weather conditions and can be installed immediately in each work area, regardless of its location. In this way, it provides a clearly differentiating and effective technological leap forward in the field of road maintenance worker safety. Some internationally known patents also offer safety solutions to road maintenance workers against unforeseen vehicle intrusions in work zones through the use of smart cones or beacons, that is, through devices associated with traffic cones, but these are systems based on different techniques, such as video image processing, infrared laser transmitter-receivers or Doppler effect, and with different functionalities. For example, the Chinese patent security system, publication number CN118736894-A, comprises a host, an intelligent traffic cone, an alarm terminal, and a handheld device. The host, equipped with a Jetson module, a camera, and radar, collects road video and radar signals to predict vehicle trajectories using target detection and tracking algorithms implemented in the Jetson module. This is combined with an intelligent cone that includes a sensor module for collecting vibration signals from nearby vehicles and a module for wireless data transmission between the host and the alarm terminal. The system warns if a vehicle's trajectory coincides with a construction lane (work zone) and if it has entered it, using different types of alarm signals. On the other hand, the Chinese utility model CN221320731-U's alert system comprises a traffic cone collision detection ring, an audible and visual alarm, and a mobile charging box, but it is based on a cloud platform. The utility model CN221261757-U's device obtains data using infrared sensors mounted on two cones located on either side of the construction zone. Korean patent KR20230114972-A describes a rubber evacuation notification cone that uses an impact detection sensor. This sensor measures the impact exerted on the cone's body, and if the measured impact value exceeds a predetermined threshold, an accident occurs. A signal is then transmitted via a communication module to a management server or user terminal.The system described in US patent 2023151567-A1 uses motion sensors associated with traffic cones near the construction zone to transmit laser pulses in the direction of approaching traffic and send a warning signal. The alert device described in patent CN115359619-A uses a cone with a distance sensor and a wireless radio frequency module, and an alarm with a wireless radio frequency module and a horn, such that when a vehicle enters a hazardous area, it transmits a signal to the alarm via the wireless radio frequency modules. The system described in patent CN114855664-A comprises a plurality of smart traffic cones that operate with infrared laser transmitters and receivers, such that when the infrared laser beam of a cone is blocked by a vehicle, a pen-shaped radio signal receiver alarm carried by the workers is activated.The smart cones of the system in utility model CN216999428-U consist of a road cone ring and a connected wireless early warning remote controller, an LED warning lamp, and an infrared detection head. The smart traffic cone of the warning system in patent CN112200989-A is also based on an infrared laser transceiver (claims). The system in patent US2022172589-A1 comprises a detection unit that can be mounted on a work zone boundary marker, such as a traffic cone (figures), which is based on the detection of moving objects using the Doppler effect. Finally, utility model CN211815744-U is also noteworthy, as it relates to a smart anti-collision cone for the automatic detection of collisions and rollovers, enabling spatial positioning and automatic on-site notification. In contrast to this state of the art, the proposed technology for detecting accidental intrusions of vehicles traveling on streets and roads in restricted work zones, using accelerometers and Artificial Intelligence that detect the unusual movement of the beacon cones, combined with immediate alerts to workers through a secure wireless communication network, represents a technological innovation, as it responds to a clear need in conservation operations, specifically to ensure the physical safety of the operators. The Artificial Intelligence component implemented in the system makes it possible to discriminate against false positives, generating more reliable and secure alerts, for example, through a more powerful audible alarm that actually puts operators in a safe situation. Furthermore, it allows the user to automatically and intelligently analyze sensor behavior patterns, profile vehicle types, detect the passage of vehicles even without physically touching the cone, and even generate statistics to improve future operations. THE INVENTION. The aforementioned intelligent system for detecting vehicle incursions into restricted road areas, delimited by traffic cones on the roadway, is based on two interconnected devices of a wireless communication network between traffic cones and operators in the application areas: - An intelligent signaling cone device, which attaches to standard road cones as an "add-on" component; and - An operator's signal receiving device, which is carried by the operators who are in the restricted area of ​​the roadway. The smart cone device consists of a polygonal prismatic body with rounded vertical edges, making it more ergonomic, with a hollow ring around its perpendicular axis, through which it is inserted into the cones, in which one of its side faces is divided into a detachable flat section compartment that houses the electronic components. At the prototype level, the device's casing is a rectangular prism, with the detachable side box highlighted in a different color. Orange for the ring body and black for the side box are ideal colors to accentuate the visual alert of the traffic cones on which they are mounted. The electronics of this smart cone device, integrated into the removable compartment, consist of the following components: - An ESP32 type microcontroller system, which integrates a two-channel radio communication module, an AI Artificial Intelligence module and a local data storage persistence module, based on a Heltec LoRA WiFi board with support for LoRA WiFi protocols, for long-range communication between cones and operators at more distant points, and support for AI; and a Nordic NRF52840 board with support for Bluetooth Low Energy protocols, for short-range communication, ensuring the connectivity of the mesh network of cones; - An MPU-6050 sensor module, a 6-degrees-of-freedom (DoF) inertial measurement unit resulting from the combination of a 3-axis accelerometer and a 3-axis gyroscope, for obtaining data to feed the AI ​​module; - A user interface, with a power button and RGB LED visual indicator of connectivity and signal quality; and, - A rechargeable battery. The local persistence module is the component responsible for storing the data handled by the device. In general, this module has the following objectives: - Storage of system configuration and parameters. - Stack of data that is pending sending but without confirmation of receipt (in the case of a cloud connection, to a server or to a destination computer). - Storage of historical data to feed the Artificial Intelligence module. - Storage of results from the Artificial Intelligence module. The operator's signal receiving device, which is carried by the operators located in the restricted area of ​​the roadway, consists of a thin box with an opening and closing lid attached to a bracelet adjustable to the user's arm, inside which is located the electronics, made up of the following components: - An ESP32 type microcontroller system with identical hardware characteristics to that of the cone device, which integrates the two-channel radio communication module, based on the Heltec LoRA WiFi board with support for LoRA WiFi protocols, for long-range communication, and the Nordic NRF52840 board with support for Bluetooth Low Energy protocols, for short-range communication, for connectivity of the mesh network of cones. - A user interface, with a power button and an RGB LED visual indicator of connectivity, signal quality, and alarm reception; an alarm button for alarm reception acknowledgment; a vibration motor that emits vibration simultaneously with the alarm message; and a KY-012 type piezoelectric buzzer that emits an audible alarm signal; and - The rechargeable battery. The AI ​​module for intrusion and impact detection integrated into the LoRA WiFi support plate of the smart cone device consists of a set of rules contained in an "Intrusion Detector" file of the local persistence module. These rules determine whether a movement within a cone equates to an impact with a probability of intrusion into the restricted area of ​​the roadway. They result from applying a "Rule Creation" labeling algorithm based on the collection and preprocessing of data from the sensor module during experiments in a controlled environment, according to the following process: i. Data collection through experiments of specific cases of vehicle impacts and intrusions into the restricted area, each experiment corresponding to a data file of a single type of movement; ii. Processing of the data files using the labeling algorithm, which assigns the corresponding labels at the times when it detects movement; iii. Generation of a data set from windows of 5 observations, in which each window is represented by a vector of statistical features: mean, standard deviation, minimum and maximum, and derived features: magnitude of acceleration and gyroscope; iv. Training a classifier with the generated dataset based on decision trees; and, v. Definition of decision tree rules, and implementation of the rules in file directly in the hardware. In both component devices of the system, the LoRA WiFi support board interconnects via GPIO lines with the Bluetooth Low Energy support board for communication of events and states, and with the user interface, implementing software developed with C language using the Arduino IDE, based on the following predefined object, variable and function files, as specific libraries: - System boot implementation file; - Implementation file for the interface and handling of the MPU6050 accelerometer sensor; - Implementation file for the collection of sensor readings to facilitate AI alarm event detection routines; - Implementation file of the AI ​​routines for the classification and detection of the different alarm events. - Implementation file for a message management network layer for the LoRA protocol; - Network quality self-perception control implementation file; - Joint expansion implementation file Also in both devices, the Bluetooth Low Energy support board interconnects via GPIO lines with the LoRA WiFi support board, for the communication of events and states, and with the user interface, implementing software developed with C language using the Segger Studio IDE, based on the following predefined variable and function files, as specific libraries: - Network level implementation file for message management for the Bluetooth Low Energy protocol, for detection and rejection of repeated messages by other nodes, and management of multiple messages to be sent by the protocol. - Implementation file for the network's self-perception quality control, with a cache table of perceived nodes and timer management for detecting node failures - Joint expansion implementation file. The main innovative and differentiating aspects of this intelligent vehicle intrusion detection system in road maintenance areas are: - New functionality for automatic and automated sending of beacon bursts upon lane intrusion detection to flood the channel and maximize alert reception. - New functionality for automated monitoring of the status of all system components: the cones periodically emit a pulse that ensures they remain operational, that the communication channel is functioning correctly, and that the receiver is operational. This pulse would be represented as a small acoustic pulse, allowing operators to know that the system is operational, and in case the receiver detects any fault, it would alert them for resolution. - Redundant communication channels - Design geared towards rapid deployment and simplified start-up, avoiding the use of specialized knowledge for its use, and streamlining its use in real situations. - Novel, highly reliable communication channel, specifically designed for critical environments in ITS (Intelligent Transport Systems) applications, capable of achieving a near 100% message reception success rate and low transmission latency. The development and implementation of this invention will offer an intelligent system for road maintenance, with the following main functionalities: - Intelligent and automatic detection of any type of access to the area delimited by the User. - Analysis of the variables collected using Artificial Intelligence technology to profile the origin of the sensor data captured by the add-on on the cones, so that the causes that produce them can be detected and false positives reduced, thus increasing the level of alert to clear vehicle incursions. - Signaling of system status, that is, the solution must indicate to the user that it is functioning correctly, if there are elements without connection, etc. - Automatic alert via acoustic signal to the operators in case of any invasion of the delimited work area. - Robust and reliable wireless communication from the smart cones to the work area. FIGURES AND GRAPHS. - The following ten figures, with images and diagrams of the components of the developed system, are included at the end of this descriptive report: Figure 1 represents a distribution of cones with an intelligent function emitting a signal (the nodes coupled to the cones are not distinguishable in the drawing). Figure 2 shows an image of the smart cone device prototype, in its preferred design embodiment, a rectangular prism-shaped piece with a central ring and a side compartment of electronics of a different color. Figure 3 is an illustration of the placement of the above device prototype on a standard cart cone, with the electronics compartment facing the restricted area. Figure 4 is a 3D model of the smart cone device prototype, and Figure 5 shows 3D models of the two disassembled prototype pieces, in which the detail of the cover where the electronics are located can be observed. Figure 6 is a 3D model of the prototype receiver operator device, consisting of an elongated box adapted to a bracelet, and Figure 7 shows 3D models of the open box and the lid. Figure 8 is a schematic of the electronics of the intelligent signaling cone device, and Figure 9 is a schematic of the electronics of the signal receiving operator device, in which it can be seen that the microcontroller system with two motherboards of both devices is the same in terms of hardware components. Finally, Figure 10 is a schematic of the AI ​​module's impact detection process. METHOD OF IMPLEMENTATION. - As can be seen in the aforementioned figures, the present system for alerting road maintenance workers that a vehicle has accidentally invaded the work area delimited by cones on the roadway, consists of a wireless communication network between the cones (1) and the workers themselves by means of two electronic devices of structural design suitable for their placement. On one side, the smart cone signaling device (2), which is the one that attaches to the road cones, made up of a polygonal prismatic body (3) with a hollow ring (4) in the center and rounded vertical edges (5), in which one of its side faces is the detachable compartment (6) that houses the electronic components: the ESP32 microcontroller system, made up of two motherboards, a LoRA WiFi support board (7) and a Bluetooth Low Energy support board (8); the MPU-6050 sensor module (9); the user interface, with power on / off button (10) and RGB LED (11); and the battery (12). On the other hand, the operator receiver device (13), which is carried by the operators located in the restricted area of ​​the roadway, consists of a thin box (14) with a lid (15) that opens and closes, attached to a wristband (16). Inside, the electronics are located, comprising an ESP32 microcontroller system with identical hardware characteristics to that of the previous element; a more complex user interface, because in addition to the power button and the RGB LED, it includes an alarm button (17), a vibration motor (18) that emits vibration simultaneously with the alarm message, and a KY-012 piezoelectric buzzer (19) that emits an audible alarm signal. The rechargeable battery is also the same. The design and development of this system architecture, including the hardware and software components, is described below. This includes the mechanisms, interfaces, and communication protocols necessary for prototype-level operation, as well as the AI ​​Module for intrusion and impact detection. 1. System architecture. -1.1. Design considerations. - The following roles are defined within the system, which generate the system architecture: The cone represents the node associated with each of the cones that delimit the restricted area of ​​the road. This node, as a component, must have a housing that easily attaches to the cone and allows access to its battery charging. Additionally, it will include certain actuator elements that can provide additional information to the operators (indicator lights). Operator, representing a track worker. Their device will be designed for comfortable wear, incorporating various actuators that allow them to communicate and receive information, such as alarms or the status of the network. At the level of role distinction, it is defined that at a "low" level there is no difference (or if there is, it is minimal) between them; that is, the software is practically identical. The same applies at the hardware level, as they will integrate the same type of components. The main difference will be the design approach to the structure and casing, as well as the possible actuators for event indication. Cone: Incorporates inertial sensors for alarm detection. It can integrate different elements into the interface to indicate certain states (light actuators). Designed to be placed on the cone, around its perimeter. Operator: They do not integrate inertial sensors. However, the interface is more complex with different actuators (light, vibration, sound, etc.). Designed to be worn primarily on the operator's forearm using a fastening strap. Furthermore, to provide greater robustness, reliability, and redundancy, two communication channels with different properties are used to ensure more reliable propagation and reception of communications. The following communication technologies are used for this purpose: LoRA: which is a low-power protocol with high reliability and a wide coverage range. Bluetooth Low Energy: which is a low-power, inexpensive and widely available protocol with a medium coverage range. Each node has this pair of communication channels, executing processes in parallel (communicating between them, if necessary, regarding certain events and states). 1.2. Communications architecture. - All the cones contain a complete radio system, integrating LoRa and Bluetooth Low Energy communication technologies. A representation of the wireless network of smart cones is shown in Figure 1. This network must have the following characteristics: Redundancy: achieved through the installation of communication technologies in all equipment. This ensures that if one element is lost or fails, alternative paths or receivers maintain the required functionality. Range: through the use of two technologies. LoRA allows a direct range to distances greater than 1 kilometer, while Bluetooth Low Energy allows efficient communication at short and medium distances, around 10-200 meters. Latency: This is achieved by using two technologies that allow for faster communication over short distances and one with slightly higher latency but a longer range. This allows receiving devices closer to the event to be notified sooner. Reliability: The use of two communication channels allows for two paths to the recipients. This increases the delivery guarantee despite a temporary failure or maintenance of one of the two channels. Additionally, they can be used to confirm detection at the receiving end and ensure with greater certainty that an emergency has occurred. Robustness: The use of both technologies ensures that, in the event of a failure of one technology, another is always available. This can also be used to detect system failures, such as when one of the channels is in fault mode for a period of time. In this way, Bluetooth Low Energy technology is employed using flooding and message broadcasting techniques across the network, progressively reaching all network elements and protecting against the isolated failure of one of them. Adjusting this broadcasting will be an important factor to consider in minimizing collisions and reducing delivery latency. Furthermore, LoRa technology enables direct communication over greater distances with a high delivery guarantee. Thus, the combination of both technologies provides a high level of fault tolerance and reliability, necessary for this type of application. The system consists of the following main states: Deploy: This is the initial state in which each node in the system begins generating and receiving Alive messages from the other nodes and builds its table, but without issuing warnings about node failures. Once the time limit for remaining in this state has expired, it returns to normal mode. Normal: which is the state in which Alive messages are still being received and generated. In this case, alerts are issued regarding self-detected node failures. These failures can be reported through the interface. Alarm: This is the state entered after an alarm situation is detected by the inertial sensor or after a message is received through a channel. In this state, any incoming messages requiring processing are ignored, messages in the outgoing message queue are removed, and alarm messages begin to be sent through both channels. Once the message transmission time has elapsed, and subsequently the Alarm state timeout has expired, the system returns to the Deploy state. Within the architecture, two fundamental behaviors must be distinguished: Architecture maintenance: which relates to the methods, procedures and processes that must be included to allow for streamlining the installation and deployment, as well as obtaining the maximum possible information from each element of the network as well as the overall set. State of emergency: which includes the processes related to the communication and propagation of an alarm (in this case the "possibility of lane invasion"). The separation between these two behaviors is total, meaning that they can evolve separately, without having (a priori) dependencies between them. 1.3. Design of the Smart Cone prototype. - 1.3.1. Format and placement. - The preferred format for the smart cone element has been the rectangular prism design with a central ring and rounded vertical edges shown in Figure 4, which is the one that best fits on the top of the road cones, minimizing the obstruction of the reflective surface. The flat-section box (6) on the side, which houses the device's electronics and is shown disassembled in Figure 5, serves as a visual reference for the operator, indicating the correct orientation of the device on each traffic cone (towards the inside or outside of the cut lane). Furthermore, in the prototype, the body with the central ring is orange, while the flat-section box is black, to reinforce and highlight the visual reference. 1.3.2. Electronic components. - Figure 8 shows the electronic component diagram of the smart cone prototype; namely: ESP32 Microcontroller: The core of the prototype is an ESP32 microcontroller, which provides processing and connectivity capabilities. This microcontroller includes a LoRa interface (7) for long-range communication and also incorporates an nRF52840 Bluetooth BT chip (8) for connectivity to the mesh network of cones. Sensors: MPU-6050 6-axis accelerometer and gyroscope (9): These sensors enable the acquisition of precise motion data, which feeds the Artificial Intelligence (AI) model designed to detect impacts. Features: - Power supply: 3-5 V (internal low-voltage differential voltage stabilization) - Communication method: standard I2C communication protocol - The chip is equipped with a built-in 16-bit AD converter and a 16-bit data output - Gyroscope range: ± 250, 500, 1000, 2000° / s - Acceleration range: ± 2, ± 4, ± 8, ± 16 g User Interface: Power button for easy control of the device. RGB LED indicator, which provides visual information about connectivity and signal quality. 1.4. Design of the alarm receiver operator prototype. 1.4.1. Format and ergonomics. - A wristband-style case design was chosen for ease of use and portability by workers during their workday. The ergonomic design ensures a comfortable and secure fit on the user's arm, allowing it to be positioned optimally for accurate haptic notification of alarms (visual, audible, and tactile). The box form factor (14) is intended to be as thin as possible, taking into account that it is a prototype, so that it can be placed both above and below the operator's layers of clothing (jackets or raincoats). 1.4.2. Electronic components. - Figure 9 shows the electronic component diagram of the receiving operator prototype, which differs from the emitting node in that the sensors are replaced by a more complete user interface; namely: ESP32 Microcontroller: As mentioned, it is the heart of the prototype, providing processing and connectivity capabilities. This microcontroller includes a LoRa interface (7) for long-range communication and also incorporates a Bluetooth BT chip (8) nRF52840 SoC for connectivity to the mesh network of cones. User Interface: Power button for easy control of the device. User interaction button, to control alarm reception recognition (turn off other active interfaces) RGB LED indicator, which provides visual information on connectivity, signal quality, and alarm reception. Vibration motor, which emits a vibration with the alarm message. Buzzer KY-012, a speaker that emits an audible alarm signal. Features - Operating voltage: 3.5 V to 5.5 V - Maximum current: 30 mA / 5 V - Resonance frequency: 2500 Hz ± 300 Hz - Minimum sound output: 85 dB at 10 cm - Operating temperature: -20 °C to 70 °C .5. Software implementation. The software implementation is based on the use of two different microcontrollers: the LoRA WiFi (7) and Bluetooth Low Energy (8) boards, which support the different communication protocols. Specifically: Heltec LoRA WiFi v3 ESP 32: which is a board with processing capabilities and support for WiFi and LoRA communication protocols. Nordic NRF52840: which is a board with support for Bluetooth Low Energy protocols. 1.5.1. LoRa Programming - Heltec LoRa wifi v3 ESP32. - This device offers greater computing capabilities. It is programmed using the Arduino IDE and a C language adapted for Arduino. Because this device is more powerful, it integrates AI support for alarm detection and manages communication with the inertial sensor. It supports the LoRa communication protocol. The device connects via GPIO lines to the Nordic NRF52840 board for event and status communication. Additionally, it has several GPIO lines for communication with the user interface. The app's design is based on the definition of several files that integrate the objects, variables, and functions. The main ones are: Smartcone.ino: This is the main file that handles system startup. It defines some general architecture constants, loads commonly used libraries, and defines some frequently used functions. Furthermore, due to the way the LoRa protocol works, the implementation of its interface must be included in this file. SensorAccelerometer.{h, cpp}: which contains the implementation that allows the interface and handling of the MPU6050 accelerometer sensor. It includes methods for obtaining readings and configuring the sensor. CaptureWindow.{h, cpp}: which contains the implementation of a class that holds a collection of sensor readings that will later be used to provide AI alarm event detection routines. FallDetectionAI.{h, cpp}: which includes the implementation of AI routines for the classification and detection of different alarm events. CHM_NET.{h, cpp}: This layer includes the implementation of a message management network layer for the LoRa protocol. It adds fields necessary for packet identification, congestion control when relays are enabled, duplicate packet management, caches of received messages, and message queues. It relieves the application layer of message sequencing control. CHM_APP.{h, cpp}: which contains the application implementation itself. It is based on a machine that cycles through defined states (Deploy, Normal, Alarm) according to the different events it detects, whether through the interface, sensors, or received messages. CHM_QOS.{h, cpp}: which contains the implementation of the network's self-perception quality control. It contains a cache table of perceived nodes and the management of the timers necessary for detecting node failures. The additional libraries used are: MPU6050 (communication with the inertial sensor), Wire (required by the MPU6050 library), Heltec_ESP32_loRA (supports LoRA communication), RadioLib (used by the LoRA protocol), SPI (required for communication with the inertial sensor via SPI protocol). 1.5.2. Bluetooth low energy programming - Nordic nrf52840.- This other device supports the Bluetooth Low Energy communication protocol. It is programmed using the Segger Studio IDE and the C programming language with various libraries, as well as a modified stack that enhances the robustness of the Bluetooth Low Energy communications. The device connects via GPIO lines to the Heltec LoRA WIFI v3 board for event and status communication. The app's design is based on the definition of several files that integrate the variables and functions. The main ones are: `chm_net_layer.{h, c}`: This layer includes the implementation of a message management network layer for the Bluetooth Low Energy protocol. It adds fields necessary for packet identification, congestion control when relays are enabled, duplicate packet management, caches of received messages, and message queues for transmission. It relieves the application layer of message sequencing control. The following depend on this layer: or chm_net_layer_cache.{h, c}: which implements the cache that allows detecting and rejecting messages already repeated by other nodes. or chm_net_layer_queue.{h, c}: which implements the management of multiple messages to be sent by the protocol. It selects the message to send at the corresponding time and updates the corresponding fields. chm_app.{h, c}: which contains the application implementation itself. It is based on a machine that cycles through defined states (Deploy, Normal, Alarm) according to the different events it detects, either through the GPIO lines or by receiving messages from the protocol itself. chm_qos.{h, c}: which contains the implementation of the network's self-sensing quality control. It contains a cache table of perceived nodes and the management of the timers necessary for detecting node failures. The libraries used are the general ones offered by the programming environment for Nordic, which are equivalent to the standard C libraries, in addition to using a custom implementation library that modifies the Bluetooth Low Energy communications stack to provide more robustness. 1.6. Interfaces. - As mentioned in the design, the most notable difference between the cone and operator roles lies in the user interfaces that allow users to indicate the various states and information generated by the system. These interfaces need refinement, although the following are currently being considered: Operator Interface: It is the most complex interface, as it must include different actuators that allow the operator to perceive the most important events, such as: Light indicators, which allow you to indicate that it is in an alarm state, and additionally with different colors or flashing schemes to communicate about the status of the self-perceived network. Vibration, to generate vibrations with sufficient intensity to be perceived, in case of being in an Alarm state. Sound, which, in the case of being in a state of Alarm, allows generating sounds that can be perceived by the user. Button, which is used to perform possible actions, such as, for example, deactivating the acoustic actuator once the alarm is perceived. Cone Interface: It's a simpler interface and will initially include: Light indicators: that allow indicating that it is in an alarm state, and additionally with different colors or flashing schemes to communicate about the status of the self-perceived network. 1.7. Architecture maintenance. - All methods, procedures, and processes related to the installation, deployment, and functional maintenance of the network of elements that enable communication of detected emergency events must be planned. In this way, the following processes can initially be determined: Deployment: the process responsible for determining whether, once installed, the system can respond adequately. In this case, it simply involves defining a reasonable timeframe to allow the generation of the network's self-perceived and crucial information. Self-perceived quality: process responsible for determining connectivity in the infrastructure autonomously, based on the information perceived in the Deploy process. Therefore, the following states are defined for verifying the status of the network architecture: 1.7.1. Deploy. - This process is performed after each element is powered on / installed, or after a system reorganization following an alarm. Each element begins transmitting Alive beacons to the system on both communication channels, which operate independently. The receiving elements insert each beacon into a network membership cache. This dynamically builds a membership table. In Deploy mode there is no timer associated with each cache entry, so if Alive packets are no longer received from a particular node, this will not cause an indicator of the self-perceived communication failure to be declared at this moment. Once the established time in this state has elapsed, it transitions to a Normal operating state. In this state, the self-perceived quality of each node is assessed. 1.7.2. Self-perceived quality. - This process is performed periodically by all elements of the infrastructure. The goal is to emit certain types of events or messages that, based on the responses, help determine the quality of the link with potential neighbors and / or nodes that make up the network. The message transmission interval must be carefully studied to find a balance between information availability and message saturation. To achieve this, and once the Deploy state has ended on each communication channel and the system is in Normal operating mode, each element periodically emits an Alive message with a TTL (Time To Live) value of 10 (although these values ​​can be adjusted). As with Deploy mode, nodes populate their system with new entries if they receive Alive messages from a new node, or they update their state with a temporary notation of the last message received. In this way, it is possible to periodically determine if all nodes in the network are transmitting their Alive messages. If a node experiences a failure or loses connection to the network, its Alive messages will not be generated, and it will be possible to detect each node that has exceeded its last message validity timer. If a node and communication channel detects an expired timer, its internal state will enter an operating mode with detected inconsistency, and this state will be communicated through the interface actuators. If communication with the missing node is re-established during this state, it will be updated in the table, and the communication channel will once again be considered reliable. On the other hand, if a node is detected as not receiving Alive messages after a certain period of time, it will be removed from the perceived membership table. 1.7.3. Timers. - One of the important aspects, which will need to be refined as the project evolves and tests and experiments are conducted, is the adjustment of the timers that affect the two aforementioned states. Specifically, the most important ones are: The Alive Interval refers to the periodic interval at which each node (and channel) transmits an Alive message. A short interval ensures greater system liveness at the cost of increased message overload on the network and higher energy consumption. A long interval will decrease the accuracy of the Alive message but will not saturate the network with messages. Adjusting this value is important to find the best balance between saturation, accuracy, and energy consumption, especially in the Bluetooth Low Energy communication channel, which operates by broadcasting and repeating messages. Max live time refers to the time frame for considering a node as failed within the system's node self-perception table. A low value will allow for a faster reaction to changes at the cost of potential false positives, while a high value will eliminate possible erroneous detections at the cost of a slower response time to failures. Delete node time, which refers to the time period used to delete a node detected as failed in the table. Adjusting these intervals will allow for fine-tuning the system's behavior. These intervals can be refined based on experience and testing of the system itself. Looking ahead, additional complex strategies could be established to improve system performance in terms of accuracy, energy consumption, and network saturation. Some proposals include: Change in transmission power values ​​according to the number of nodes. Number and period of repetitions per message, which can be based on Gaussian techniques, etc. Changes to decision window timers. 1.8. State of emergency event. - This refers to the system's operation, once the components are deployed, which detects emergency events involving the potential entry of a vehicle into the work area. This process must also operate autonomously, without explicit operator involvement, and establish a system state that allows for a response to such events. In this case, each element of the infrastructure, including the operators, integrates a system comprised of the hardware that supports the proposed communications architecture. Additionally, if it is a cone-type or signaling node, it will also integrate the hardware necessary for detecting a possible encroachment on the construction lane, as well as the indicator lights or other indicators used to communicate the status (related to the processes involved in maintaining the architecture). Furthermore, if it is a node carried by an operator, it will integrate the equipment that allows the operator to be notified of the detected emergency in some way (audible, visual, vibration, etc.). The operating mode is as follows, depending on whether the node (cone) detects the fall directly through the inertial sensor, or detects it by listening to alarm messages from other nodes. In this case: a) When a node detects a potential emergency state through the inertial sensor, it goes into alert state and periodically sends a warning of this state through both channels for a set time: If it's a LoRa channel, the message will be received by all devices. The alert detection by the inertial sensor is propagated via a signal to the Bluetooth Low Energy channel. If it is a Bluetooth Low Energy channel, the messages are disseminated through the rest of the nodes until they also reach the destination nodes. b) When a node detects a potential emergency state through a message generated by another node, the node goes into alarm state and periodically sends a warning of this state through both channels for a set time: If it's a LoRA channel, the message will be received by all devices. The Bluetooth Low Energy channel is notified that an alarm has been received via the LoRA channel. If it's a Bluetooth Low Energy channel, the messages are disseminated through the other nodes until they reach the destination nodes. The LoRa channel is notified that an alarm has been received via the Bluetooth Low Energy channel. c) In all cases A minimum emergency state duration is established before guaranteeing that the infrastructure returns to a stable and operational state. This duration must be at least twice the alarm state duration. The system returns to a Deploy state to ensure connectivity again. Different repetition interval values ​​can be set, depending on whether it is the original detector node (alarm detection via inertial sensor) or by reception from other neighbors. Additionally, other heuristics for dissemination control can be established based on data collected during maintenance processes, such as adjusting timers, repetition values, or transmission power based on the number of neighbors, signal quality values, etc. 2. AI module for intrusion and impact detection. The Artificial Intelligence (AI) module embedded in the cone device must be able to analyze the system's sensor data to more reliably estimate the potential intrusion of a vehicle into the delimited work zone. This reduces false positives generated by other movements due to variables such as traffic proximity, accidental drops of cones, and handling by personnel. To achieve this, it relies on N:1 and N:K machine learning models that predict the cone's state relative to its movement, identifying false positives based on source data, involved variables, and the application of deep learning techniques such as auto-encoders, LSTM, and hybrid anomaly detection methods on derived metrics. The AI ​​module consists of two main components: Rule creation: Algorithm that allows calculating and extracting the rules that determine if a movement in the cone is equivalent to an impact with a probability of intrusion into the work area. Intrusion Detector: A set of rules, resulting from the previous step, that are implemented in the microcontroller (hardware) to detect impacts and intrusions. Therefore, the rule creation process is based on applying an algorithm from the collection and preprocessing of sensor data during an experiment in a controlled environment. Figure 10 details the process by which the impact detection system is built. First, experiments are conducted to collect data on the events to be detected, such as vehicle impacts and intrusions into the work area. Each experiment corresponds to a file containing a single type of movement. These files are processed using a labeling algorithm that assigns the corresponding labels at the moments when movement is detected. Subsequently, a dataset is generated from windows of 5 observations, where each window is represented by a feature vector. The features include statistical measures such as mean, standard deviation, minimum, and maximum, as well as derived features (e.g., acceleration and gyroscope magnitudes). A summary of the previous window is also included to reflect the evolution of the motion. Using this dataset, a decision tree-based classifier, chosen for its explainability, is trained. A hyperparameter search is performed using cross-validation, with accuracy as the metric, and the dataset is divided into 70% for training and 30% for evaluation. Finally, the decision rules are extracted from the tree, allowing the model's internal logic to be translated into simple (if / else) conditions. These rules are then implemented directly in the hardware to achieve real-time intrusion and impact detection.

Claims

1. Intelligent system for detecting vehicle incursions into restricted road zones, delimited by traffic cones on the roadway, based on ITS technologies for intelligent transportation systems for road safety, characterized by two interconnected devices of a wireless communication network between traffic cones (1) and operators in the application areas: - An intelligent signaling cone device (2), which attaches to standard road cones as an "add-on" component, consisting of a polygonal prismatic body (3) with a hollow ring (4) around its perpendicular axis, through which it is inserted into the cones, and rounded vertical edges (5), in which one of its lateral faces is divided into a detachable flat-section compartment (6) that houses the device's electronics, consisting of the following components: An ESP32 type microcontroller system,integrating a two-channel radio communications module, an Artificial Intelligence (AI) module, and a local data storage persistence module, based on a board supporting LoRA WiFi protocols (7) for long-range communication and AI support, and a board supporting Bluetooth Low Energy protocols (8) for short-range communication; an MPU-6050 sensor module (9), a 6-degree-of-freedom (DoF) inertial measurement unit resulting from the combination of a 3-axis accelerometer and a 3-axis gyroscope, for obtaining data to power the AI ​​module; a user interface with an on / off button (10) and an RGB LED (11) visual indicator of connectivity and signal quality; and a rechargeable battery (12). - A signal receiver (13) operator device, which is carried by the operators located in the restricted area of ​​the roadway,consisting of a thin box (14) with a lid (15) that opens and closes, attached to a bracelet (16) adjustable to the user's arm, inside which is located the electronics of the device, consisting of the following components: An ESP32 type microcontroller system with identical hardware characteristics to that of the previous element, which integrates the radio communication module for two channels, based on the board with support for LoRA WiFi protocols (7), for long-range communication, and the board with support for Bluetooth Low Energy protocols (8), for short-range communication; a user interface, with the power button (10) and the RGB LED (11) visual indicator of connectivity and signal quality and alarm reception, an alarm button (17), for acknowledgment of alarm reception, a vibration motor (18) emitting vibration simultaneously with the alarm message,and a KY-012 type piezoelectric buzzer (19) emitting an audible alarm signal; and the rechargeable battery (12).

2. An intelligent vehicle intrusion detection system in restricted road zones, according to claim 1, characterized in that the polygonal prismatic body (3) that forms the intelligent cone device is a rectangular prism with a detachable side compartment (6) highlighted in a different color.

3. An intelligent vehicle intrusion detection system in restricted road zones, according to claim 1, characterized in that the AI ​​module for intrusion and impact detection integrated into the LoRA WiFi support plate (7) of the intelligent cone device consists of a set of rules contained in an "Intrusion Detector" file of the local persistence module, rules that determine whether a movement within a cone equates to an impact with a probability of intrusion into the restricted road zone.resulting from applying a "Rule Creation" labeling algorithm to the collection and preprocessing of data from the sensor module, according to the following process: i. Data collection through experiments of specific cases of vehicle impacts and intrusions into the restricted area, each experiment corresponding to a data file of a single type of movement; ii. Processing of the data files using the labeling algorithm, which assigns the corresponding labels at the moments when movement is detected; iii. Generation of a dataset from windows of 5 observations, in which each window is represented by a vector of statistical characteristics: mean, standard deviation, minimum and maximum,and derived characteristics: magnitude of acceleration and gyroscope. iv. Training a classifier with the generated dataset based on decision trees; and v. Definition of decision tree rules, and implementation of the rules in a file directly in the hardware.

4. Intelligent vehicle intrusion detection system in restricted road areas, according to claim 1, characterized in that the LoRA WiFi support board (7) interconnects via GPIO lines with the Bluetooth Low Energy support board, for the communication of events and states, and with the user interface, implementing software developed with C programming language,Based on the following files implemented as specific libraries: - System boot implementation file; - Implementation file for the interface and handling of the MPU6050 accelerometer sensor; - Implementation file for the collection of sensor readings to be provided to the AI ​​alarm event detection routines; - Implementation file for the AI ​​routines for the classification and detection of the different alarm events; - Implementation file for a message management network layer for the LoRA protocol; - Implementation file for the network quality self-perception control; - Implementation file for the overall expansion.

5. Intelligent vehicle intrusion detection system in restricted road areas, according to claim 1, characterized in that the Bluetooth Low Energy support board (8) interconnects via GPIO lines with the LoRA WiFi support board for the communication of events and states.and with the user interface, implementing software developed in C based on the following files implemented as specific libraries: - Network layer implementation file for message management for the Bluetooth Low Energy protocol, for detection and rejection of messages repeated by other nodes, and management of multiple messages to be sent by the protocol. - Network quality self-perception control implementation file, with a cache table of perceived nodes and timer management for detecting node failures. - Overall expansion implementation file.

Images