AISCS: Autonomous, Intelligent Sign Control System using Wireless Communication and LED Signs for Rural and Suburban Roads

Depósito de Investigación de la Universidad de Sevilla https://idus.us.es/ This is an Accepted Manuscript of an article published by IEEE: “A. Torralba, J. P. García-Martín, J. M. González-Romo, M. García-Castellano, J. Peral-López and V. Pérez-Mira, "An Autonomous, Intelligent Sign Control System Using Wireless Communication and LED Signs for Rural and Suburban Roads," in IEEE Intelligent Transportation Systems Magazine, vol. 14, no. 2, pp. 115-128, March-April 2022, doi: 10.1109/MITS.2021.3049375”. © 2022 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.” > IDENTIFICATION NUMBER: ITSM-19-07-0144 < Abstract—The number of road accidents still remains too high nowadays. Different initiatives are being taken to reduce them, whether acting on the vehicle, the driver, or the infrastructure. It is well known that most incidents occur in rural and suburban areas, which have low, unattended traffic. This work presents AISCS, an intelligent, robust, and autonomous sign control system based on wireless communication and LED signs. The controlled toggling of the illuminated signs alerts drivers at specific accidentprone locations on the route. The architecture of AISCS is based on video cameras, which detect traffic conditions, and on sensing units that monitor environmental variables that can impact traffic. These are connected to a local controller with a decision-making logic with two layers of intelligence. The upper layer uses heuristic rules to classify the scene, while the lower layer uses a fuzzy logic controller to act on autonomous LED signs. The results obtained from two demonstrators, which have been in operation for more than two years in rural and suburban areas, respectively, are also shown. AISCS improves ongoing traffic safety and smoothness compared to traditional non-highlighted traffic signs, reducing power consumption by more than 90% against an always-on system. Index Terms— Intelligent Transportation Systems, Fuzzy Logic Control, LED signs, Road Traffic, Smart Signs, Traffic Sign Control, Wireless Sensor Networks I. INTRODUCTION ESPITE the efforts made by different administrations and automobile associations, the rate of road accidents in the Western world still remains too high. The rate of decline in fatal road accidents in the European Union (EU) has gone from 43% during 2001-2010, to only 2% in 2016-2017. This fact has led to the "Valleta Declaration on Road Safety", signed in 2018 by Transport Ministers of the EU [1], [2], which has paved the way for a new reduction target of 50% in serious injuries between 2020 and 2030. Of special importance, 55% of road fatalities in Europe in 2017 occurred on rural roads, which also show a seasonality higher than in urban areas. As stated in [1], this is probably because changes in daylight hours and weather conditions have This work was supported in part by the Consejería de Fomento y Vivienda, of the Andalusian Regional Government, and by the European Regional Development Fund, under Grant G-GI3002/IDIO (SRA/ERDF EU). A. Torralba, J.P. García-Martín, and J.M. González-Romo, M. García- Castellano, and J. Peral-López, are with the Electronic Engineering a higher impact on those areas. Among the actions proposed by the EU to reduce road accident rate, improvements in infrastructure stand out [3], especially those that affect the visibility and readability of intersections, which highlights the importance of improvement in traffic signs and in the illumination of intersections on secondary roads [4]. The actions adopted to reduce the high number of fatalities that occur in accidents in suburban and rural areas are normally limited to works that improve their accessibility and visibility. Although numerous proposals can be found in the literature to increase safety by the introduction of active signs, they are seldom used due to the lack of power supply and good communication links in remote areas, and to installation costs. Recent introduction of low-power LED signs, the development of wireless sensor networks, and the use of autonomous control systems allow the development of cost-effective active sign control systems, which must be validated and tested in real traffic conditions. This paper presents an Autonomous Intelligent Sign Control System (AISCS) based on active LED signs, for use in rural and suburban areas. Compared to other related works, AISCS has the following relevant features: - A two-level decision-making logic: the upper one deals with scenario classification, while the lower one uses fuzzy inference to act on LED traffic signs. - Non-intrusive: none of AISCS elements is located on or under the pavement; therefore, their commissioning or repair does not imply traffic detention [5]. - Self-powered: which avoids the need to run a power cable from the nearest power supply. - Wireless communications: it is not necessary to carry out canalization works for their deployment or replacement. Section II reviews the state of the art. Section III shows AISCS functionalities and selected Use Cases. Section IV, V, VI, and VII present the logical architecture of AISCS, the communication network, the local processing layer, and its physical implementation, respectively. Section VIII discusses the main AISCS innovations. Section IX shows the results Department, and V. Pérez-Mira with the Chemical Engineering Department, Universidad de Sevilla (Spain), email: [email protected], {jpgarcia, jgonzalez, mgarcia}@gie.us.es, and {jperal1, vperez4}@us.es, respectively. AISCS: Autonomous, Intelligent Sign Control System using Wireless Communication and LED Signs for Rural and Suburban Roads Antonio Torralba, Senior Member, IEEE, Juan Pablo García-Martín, Juan-Manuel González- Romo, María García-Castellano, José Peral-López, and Ventura Pérez-Mira D > IDENTIFICATION NUMBER: ITSM-19-07-0144 < obtained in the selected Use Cases, where two prototypes of AISCS have been in operation for more than two years. Finally, in Section X some conclusions are drawn. II. RELATED WORK Many proposals to increase road safety and to improve accident prevention can be found in the literature. Advances in communication technologies and control algorithms are also giving rise to new methods of traffic monitoring and control. Drivers can be warned of possible risks through road signs. However, due to driver distractions, as well as low visibility conditions, conventional signs may go unnoticed. Guidelines on suitable lighting requirements for bollards and traffic signs can be found elsewhere [6]. Recently, especial attention has been given to intelligent lighting control on road lanes and to retroilluminated signs. In [7], a sign controller powered by solar panels is presented. Different procedures are established for day operation (when the battery charges through the solar panel) and night operation (when the battery discharges). An intelligent street lighting system based on Zigbee is shown in [8], where some results are presented at simulation level, with an emphasis on intersections. In [9], a new generation of wireless digital signs is introduced, according to the V2I paradigm, where conventional signs are replaced by intelligent poles with WiFi connectivity. As vehicles approach a pole, they receive notifications and warnings corresponding to one or more vertical signs. This approach is similar to some others, where cars read a Radio-Frequency Identification (RFID) tag attached to road signs through an onboard RFID reader, although a WiFi network provides greater coverage and transmission capacity. Other examples of wireless communications and Wireless Sensor Networks (WSNs) for intelligent signs appear in [5], [10]-[12], where Zigbee and WiFi are the selected standards. The system proposed in [11], by which roads signs transmit FM commercial radio signals to vehicles, deserves a special mention. Visual and audible alarms are generated by taking advantage of their onboard radio receiver. It is a low-cost solution, and it is not affected by regulatory impediments. In [5], an example of traffic monitoring at the freight village of Turin is shown. Regarding traffic signs, [13] reviews self-adaptive methods with up to 5 levels of intelligent decision-making. It surveys advanced control systems such as SCOOT [14], RHODES [15], and AFT [16]. Different predictive algorithms and green-wave strategies are also used for traffic control and monitoring. One of the main goals is to minimize the stop and delay times at intersections [17]-[18]. Other systems, such as [19], use image processing, which generates notifications to the operators of a traffic control system for Smart Cities from the images captured by traffic control cameras, Other intelligent systems based on video processing can be found in [20]-[22]. In [21], Unmanned Aerial Vehicles (UAVs) are used with aerial video cameras. Other solutions offer other types of driver assistance, such as the one shown in [23], which presents a Driver Advisory System (DAS) where the lane separation lines are illuminated when the presence of a vehicle is detected, and signs are also automatically detected and recognized from inside the vehicles. Table I summarizes the review of the main related works. III. SYSTEM FUNCTIONALITIES & USE CASES A. System Functionalities AISCS is an intelligent system for LED road signs based on the non-intrusive installation of sensors and actuators on the roadside, with the aim of reducing accident rate in rural or suburban areas, with special attention to unregulated intersections. For this reason, AISCS measures: - Traffic conditions (number and speed of vehicles approaching the selected area); - Road conditions (the presence of pedestrians and animals near or on the road, and the presence of fixed and mobile obstacles); - Environmental conditions (luminosity, humidity, rainfall, and temperature). This information arrives in a safe and reliable way to a Local Controller (LC), which: - Stores the measured data and produces statistics; - Processes the incoming information, to independently act on each of the LED signs (activating its light, with a selectable luminous intensity and toggling frequency); - Communicates with the Central Controller (CC) for different functions, such as alarm management, reporting of system state and statistics, maintenance tasks, remote programming and / or system update. The main functionalities of AISCS are the detection and signaling of a) an excessively high approaching speed of a vehicle, b) the presence of other vehicles or obstacles that obstruct traffic, and c) the presence of other elements of risk (pedestrians, bicycles, skates, animals, etc.) near the selected area. To allow its massive deployment, the system has low costs of acquisition, installation, operation, and maintenance. Other specific performances of the system are: - Autonomy: The capacity of the batteries for each of the signs and sensing elements is designed to ensure continuous operation; - Flexibility: System software easily incorporates new sensors and actuators, which expand its capacity to act in future cases beyond its present scope (for example, under snow conditions); - Resilience: The system is able to detect a malfunction in one or some of its components, and re-adapt its decision rules to ensure road safety; - Scalability: New signs and sensing elements are incorporated to AISCS, with the simple addition of a wireless node within the WSN coverage; - IPv6 connectivity in the WSN nodes, so that they can be directly integrated into the IP network. > IDENTIFICATION NUMBER: ITSM-19-07-0144 < B. Use Cases AISCS has been designed to help drivers on rural and suburban roads. Two Use Cases, representative of those scenarios, have been considered. The first one (Suburban-Area Use Case) considers an intersection between the access road in a small town and a secondary road. There is a bus stop near the intersection, which implies frequent pedestrian traffic. In this use case, specific AISCS objectives are the detection and signaling of a) pedestrians in the crosswalk and b) bus lane occupation. In the second one (Rural-Area Use Case), a straight section of a secondary road is monitored in a remote area where the crossing of stray cattle is sometimes detected. Specific objectives are then the detection and signaling of a) an animal or mobile obstacle crossing the road, and b) a moving agent close to the road sidelines. IV. LOGICAL ARCHITECTURE The logical architecture of AISCS is shown in Fig. 1. This is a distributed system, where LCs are managed and supervised from a CC, located at the facilities of the concessionary company in charge of the exploitation and maintenance of the road infrastructure. The Data Acquisition Layer (DAL) of the LC receives incoming data from video-cameras and environmental sensors (through their corresponding Sensors Controllers - SeCs) deployed in the region of interest, which the pre-processes and stores in an internal data base. The Remote Communication Layer (RCL) maintains a wireless link by using a 3G/Universal Mobile Telecommunications System (UMTS) with the remote CC. In the Local Processing Layer (LPL), the Traffic Condition Subsystem (TCS) uses an internal model to determine the environmental variables (low visibility, rain, etc.) and traffic variables of interest (no vehicles, pedestrians in the crosswalk or near the road, etc.) to AISCS. Their values, represented as membership functions, are the inputs to a Fuzzy Logic Controller (FLC), which is part of the Decision-Making Logic (DML) that determines the actions to be carried out on the Sign Controllers (SiCs), which drive one LED sign. The Local Security Layer (LSL) deals with security issues in the LC. On top of the hierarchy, the Local Management Layer (LML) in Fig. 1, is in charge of 1) commanding the activation and deactivation of the in-field actuators (LED signs), 2) the supervision of the correct functioning of sensors, actuators, and communication network, 3) the management of local alarms and emergency situations, and 4) the management of configuration parameters, including sensors and actuators, as well as the updating of the decision rules of the DML. In the CC, the Data Integration Platform (DIP) periodically interacts with the LCs, storing global operating data. It manages the interface with remote LCs via the mobile network (using 3G/UMTS), with the system operator (using the keyboard and displays), and with other services and applications in the CC (not shown in Fig. 1). TABLE I A REVIEW OF AISCS RELATED WORK Category Reference Year Application Communication Technology Mobility Traces Autonomous Smart Signs [9] 2019 Design of wireless digital traffic sign post WiFi (IEEE 802.11g) Design [7] 2015 Design and implementation of solar powered LED road traffic sign - Laboratory test [11] 2015 Design and implementation of traffic alert system using intelligent poles Global System for Mobile communications (GSM) and commercial FM Outdoor test [23] 2018 Proposal of a smart Driver Advisory System (DAS) providing lighting for road lanes - Datasets & outdoor test Intelligent Streetlighting at intersections [20] 2014 Use of video processing in smart light switching, and traffic calculation methods - Outdoor test [8] 2018 Design of remotely-controlled intelligent luminaires through simulations ZigBee Simulation Wireless control systems for ITS [10] 2013 Implementation of a traffic control system for intersections based on WSN ZigBee Design [5] 2013 Design of a traffic monitoring system in a logistic center based on WSN General Packet Radio Service (GPRS) Outdoor test Video acquisition for traffic management [22] 2016 Design of estimation methods for automotive camera rate control Ethernet Simulation [21] 2017 Design of a traffic monitoring system using Unmanned Aerial Vehicles (UAVs) - Outdoor test [19] 2017 Design of a traffic control system with video surveillance - Simulation & laboratory test Predictive algorithms and fuzzy logic control [17] 2013 Design of a traffic signal control system with V2V and V2I communications Society of Automotive Engineers (SAE) J2735 Microscopic simulation [18] 2012 Adaptive traffic light control system utilizing ad-hoc vehicular communications network Wireless (not specified) Simulation > IDENTIFICATION NUMBER: ITSM-19-07-0144 < 4 The Central Processing Layer (CPL) processes data coming from the LCs and retrieved by the DIP, in order to 1) compute system statistics, 2) supervise LCs operation, and 2) update their actuation rules, according to the experience acquired during the operation of the system. The CPL also hosts services and applications to build traffic models and to operate fuzzy controllers in remote LCs. The Central Management Layer (CML) is in charge of the global management of AISCS, including the system configuration and update. The CML is also in charge of handling global alarms and emergency situations. Finally, the Central Security Layer (CSL) deals with security issues in the CC, including identity management, authorization, data privacy, and integrity. V. AISCS COMMUNICATION NETWORK Fig. 2 shows the communication network in AISCS. A WSN, based on the IEEE 802.15.4 standard, links a set of sensor nodes deployed in the region of interest. This standard has been chosen as a balance between cost, coverage (several hundred meters), and speed (up to 250 kbps). IPv6 over Low power Wireless Personal Area Networks (6LoWPAN) specification is employed as a lightweight adaptation layer that allows IP datagrams to be transmitted on top of the IEEE 802.15.4 stack. For the WSN, a mesh topology has been chosen, where every full-functional device can act as a router. It gives multiple paths from source to destination, providing a high redundancy in the communication link. In AISCS, every sensor or actuator node is an End Device (ED). In the LC, a Border Router (BR) acts as the WSN coordinator. A sensor node is responsible for measuring some environmental variables, namely: luminosity, temperature, rainfall, and humidity. All the sensors attached to a sensor node, including the wireless terminal, conform a Sensing Unit (SeU), which sends periodic samples to the BR. Just in the same way, a LED sign and its corresponding wireless terminal is termed a Signaling Unit (SiU), which also receives periodic frames from the BR. Through its BR, the LC Unit (LCU) collects the information coming from in-field elements in AISCS, processes the received information, and sends system and control commands to the signs through their wireless terminals. Fig. 1. AISCS’s logical architecture and simplified flowcharts. > IDENTIFICATION NUMBER: ITSM-19-07-0144 < 5 The real-time Contiki Operating System (OS) [24]-[27] has been chosen for the wireless terminals. The 6LoWPAN adaptation layer [25] allows IP addressing, which is one of the requirements of AISCS. Each wireless device has a unique identifier of 16 bits (Network ID Device). In turn, the wireless nodes have an IPv6 network address (128 bits) for routing purposes. An Image Processing Unit (IPU), serving a set of video cameras, is also part of the LC. It is attached to the LCU by an Ethernet connection. Finally, the LC and the CC communicate through a 3G/UMTS connection using the network of a telecommunication carrier. VI. AISCS LOCAL PROCESSING A. Traffic Condition Subsystem (TCS) The TCS in the LC is in charge of the creation and classification of traffic scenarios. To this end, it constantly updates the environmental variables through the WSN and uses up to three cameras attached to an IPU to monitor the region of interest. Night vision cameras are used when the region of interest is not permanently illuminated. The IPU is able to detect incoming vehicles and their speed, as well as mobile and fixed obstacles, both on the road and in regions close to the road sidelines. To this end, the camera Field of View (FoV) is clustered in different areas of interest. Fig. 3(a) shows the areas of interest: “approaching area for vehicles”, “pedestrian crosswalk”, “bus stop”, and “road sidelines” for the Suburban-Area Use Case. The IPU monitors these areas of interest in order to detect the presence of moving vehicles or pedestrians and fixed or moving obstacles. These regions are defined in an eXtensible Markup Language (XML) file by their coordinates in the camera FoV, although cameras periodically refine their position (identifying contours in captured real images) in order to remove inaccuracies due to the unavoidable small changes in the camera position with time. Fig. 3. Suburban-Area Use Case: a) Field of View of one of the cameras attached to the LC, where the areas of interest are highlighted. One pedestrian close to the pedestrian crossing and one vehicle approaching the area of interest from the upper part of the figure can be observed. In addition, views of b) the “bus stop occupied” and c) the “pedestrian crossing” signs are shown. Fig. 2. AISCS communication network and physical implementation. Two Use Cases (suburban and rural areas, respectively) are considered in the figure. > IDENTIFICATION NUMBER: ITSM-19-07-0144 < 6 Fig. 3(a) shows a case where a pedestrian is close to the road and a car approaches the crosswalk at a low speed. An enlarged view of the top area of the image is superimposed on the figure. From the images captured and processed by the IPU, the TCS selects labels for a set of variables in an internal database to create the scenario. In this case, the selected labels for the eight variables of the FLC are: a) Variables relative to the area of interest. Vehicle position: “approaching”, Vehicle speed: “low”; Obstacle position: “no obstacle”; b) Variables relative to the pedestrian crosswalk: Pedestrian position: “near”; c) Variables relative to bus stops: Bus: “no bus”; d) Variables relative to environmental conditions: Luminosity: “high”; Rainfall:”no rainfall”, Temperature-humidity conditions: “good”). Note that more than one camera may be necessary to get a good view of the area of interest. Three cameras were used in the demonstrator in the Suburban-Area Use Case. B. Decision-Making Logic (DML) The smart DML in AISCS consists of two layers. The upper one is a supervisory layer that controls the flow of decisions. Based on heuristic rules, it detects the operating situation of the system: off-line (installation, maintenance, test), paused, normal operation, or alarm situation. It also supervises the operation of the lower layer, which has a FLC of the Takagi- Sugeno-Kang (TSK)-type [28]-[30]. The label assigned to every variable in the TCS is now translated to its corresponding membership function, which is an input to the FLC. By applying inference rules, a set of FLC outputs identify risk situations and trigger a sequence of events, which correspond to the activation or deactivation of the different LED signs, including the intensity and toggling frequency of their lights. In the case of the Suburban-Area Use Case, 15 different sequences have been programmed; eight of them appear in Table II. By way of illustration, Table II also includes a simplified version of the IF-THEN rules of inference. For the sake of simplicity, only two (Luminosity and Rainfall), out of 8, input variables, have been selected for this example. Three possible values for the LEDs luminous intensity I (I0, I1 and I2) and toggling frequency f (F0, F1 and F2) have been defined. Depending on luminosity range (High – H, Medium – M, Low – L or Zero - Z) and measured rainfall (Yes – Y, No – N), the FLC determines a sequence of values for the luminous intensity and toggling frequency of the LED signs. The * superscript in the sequence symbol indicates a risk situation. For the sake of simplicity, in Table II, each sequence 𝑆!" , corresponds to a predefined value of intensity and frequency. The processing in the FLC in both demonstrators is more involved, with programmable output sequences for every rule, whose corresponding intensity and frequency values are tuned in laboratory, during the training phase of the FLC. In this way, the number of the FLC rules do not exponentially scale with the number of variables. Once a sequence has been triggered, the DML upper layer takes the corresponding decisions. At any time (even within a sequence), the upper layer can re-activate the FLC with new inputs in order to trigger a new sequence of events, if a significant change in traffic conditions is detected. The decision rules of the FLC, as well as the sequences of events triggered by the DML can be read, re-programmed, or updated from the CC, at any time. It is necessary to emphasize that, once it is lit, an LED sign flashes a minimum of 60 s, in order to maintain the driver's attention for a sufficient time. Fig. 3(b) and 3(c) show the LED signs deployed in the Suburban- Area Use Case. C. AISCS: A distributed system AISCS is a distributed system because it has been designed to allow autonomous operation of the LC, even without communication with the CC or a remote server, so that cellular coverage is not strictly necessary. Note that the LC communicates with the Sensor and Sign controllers through a wireless personal area network, which is a private network that does not require cellular coverage. In any case, if cellular coverage is available, the LC can additionally exchange information with the CC, so that the rules of inference of the DML can be improved with the experience and the CC can monitor the remote operation of the LC. In future versions of AISCS, a periodic update of the LC database using some kind of "walk-by" technique (that is, by installing a CC instance in a vehicle that periodically visits those LCs which are in remote locations) will be considered, in order to achieve full AISCS functionality, even in remote areas without cellular coverage. VII. PHYSICAL IMPLEMENTATION This Section describes the implementation of the main AISCS elements. A. Wireless Terminal and Border Router (BR) The wireless terminal is termed the Communication Main Board (CMB). It uses the low-power Microcontroller Unit (MCU) TI MSP430F5438A, and contains a TI CC1125 radio transceiver to be used in the 868 MHz (Europe) or 900 MHz (USA) band, Fig. 4(a). TABLE II SIMPLIFIED FLC INFERENCE RULES. Luminosity Z L M H Rainfall N 𝑆!"! 𝑆!"! 𝑆"!! 𝑆#!! Y 𝑆## ∗! 𝑆## ∗! 𝑆"# ∗! 𝑆"#! IF (Luminosity is M AND Rainfall is N) THEN 𝑆"! (I = I1, f = F0) IF (Luminosity is L AND Rainfall is Y) THEN 𝑆## ∗ (I = I2, f = F2) … > IDENTIFICATION NUMBER: ITSM-19-07-0144 < 7 The CMB is the only hardware component of the BR. The BR is linked to its LCU by means of the UART interface using Serial Line IP (SLIP). This protocol is frequently chosen when dealing with constrained embedded systems because of its small overhead and reduced complexity [25], [27]. B. Sensor Node The sensor node is part of the SeU. At hardware level, it consists of two boards: the already cited CMB and a Signal Conditioning Board (SCB), which is intended for input signal conditioning. Fig. 4(b) shows the picture of a sensor node, where both cards are stacked. The CMB is in the lower level, and the SCB in the upper one. Note that, even though every sensor node has the same CMB, different SCBs have been designed for every sensor type. The Trend Control LLO has been used to measure luminosity, the Vaisala HMP110 to measure temperature and humidity, and the Honeywell LLE101000 to detect rainfall. Regarding the firmware, application software manages the attached sensors on top of the Contiki OS. By default, the measurement period for sensors is 120 s. To reduce out-of-band noise, the application software performs a simple low-pass filtering, by averaging the received samples. C. Actuator Node and LED signs The actuator node is part of the SiU and receives commands from the LC to start/stop blinking. Once again, it is a composition of a CMB and an ancillary board, the so-called Sign Activation Board (SAB), which has been especially designed for AISCS. Fig. 4(c) shows the actuator node, where the CMB is in the lower level, and the SAB in the upper level. The SAB provides a voltage of 0 V (light off) or 12 V (light on) to drive a Pulse-Width Modulated (PWM) signal generator, whose frequency and average values determine the light frequency and intensity, respectively. With respect to LED signs, each one has its corresponding actuator node. Table III lists the characteristics of the LED signs used in AISCS demonstrators. The instantaneous power consumption depends on the number of LEDs, reaching 860 mW in the worst cases. D. Local Controller (LC) Each LC contains the BR, the LCU and the IPU. The LCU has been implemented on a BeagleBone Black (BBB) card, with an ARM Cortex-A8 core and Ubuntu operating system. This platform contains an Ethernet port to communicate with the IPU, and a 3G/UMTS terminal to communicate with the Remote CC (Fig. 2). In our case, the IPU contains a VBOX3200 embedded computer. E. Central Controller (CC) In its present implementation the CC also uses the same BBB card as the LCU to build most of the DIP. It is attached to a Personal Computer that implements the rest of the CC, including the operator interface. In future versions, a cloud server can be programmed or linked to the CC to take profit from the computing power of a cloud server to deal with (a) (b) (c) Fig. 4. WSN devices: (a) Communication Main Board (CMB) and Border router (BR), b) Sensor node, (c) Actuator node CMB SCB CMB SAB TABLE III HOMOLOGATED LED BLINKING SIGNS USED IN AISCS DEMONSTRATORS Sign Shape Dimensions LEDs DC Supply Voltage (Vdc) Maximum Power Consumption (mW) Image Pedestrian Crosswalk Square 60cm x 60cm x 5cm 96 Red LEDs Luminous Intensity: Up to 7,2 cd 12 600 Bus Lane Rectangle 60cm x 90cm x 5cm 110 Red LEDs Luminous Intensity: Up to 7,2 cd 12 780 Farm Animals Triangle 90cm x 90cm x5cm 120 Red LEDs Luminous Intensity: Up to 7,2 cd 12 860 > IDENTIFICATION NUMBER: ITSM-19-07-0144 < 8 complex cases. The cloud server could take the responsibility for the Decision-Making Logic in the LC, as far as the communication link is not a bottleneck. Even the EDs might forward requests to this server through their IPv6 interface. Constrained Application Protocol (CoAP) or Message Queue Telemetry Transport (MQTT) could be chosen for these connections. VIII. AISCS MAIN INNOVATIONS Flashing LED signs have been commonly used on roads in many countries around the world, but they are usually comprised of an operator or a real-time clock that activates or deactivates the lighting of every sign. On the contrary, in AISCS, an intelligent system makes decisions regarding the illumination of signs based on the measurement of: - Real-time traffic condition, obtained from the information provided by a set of video cameras; - Environmental measurements that affect driving, obtained from specific sensors deployed in the area of interest. From this information, AISCS has the ability to warn drivers by means of the flashing time, as well as luminous intensity and toggling frequency of vertical LED signs. In addition, AISCS is an autonomous system that does not require an external power supply, and its elements communicate using a wireless link. Therefore, it is easy to install and does not require civil work. The combination of these features makes the installation and maintenance costs of AISCS significantly lower than those of a conventional sign control system. One of the most innovative features of AISCS is the employment of an intelligent decision-making logic in the LC. Based on heuristic rules, and according to traffic and ambient conditions, it detects the operating situation of the system, and supervises the operation of a fuzzy logic engine that acts on LED signs. The decision rules are regularly updated by the CC, which learns from the experience gained with the operation of the system. AISCS can also contribute to the V2I interaction. In this sense, some tests have been carried out for the integration of AISCS in a system for the automatic inventory of traffic signs described in [31]. In this way, updated information is available about the location, date of installation, and present state of AISCS elements, by simply attaching a passive RFID tag. Furthermore, installing an RFID reader in the vehicle would also increase the intelligence of the system, and this will be the objective of future projects. IX. EXPERIMENTAL RESULTS AND DISCUSSION Two AISCS demonstrators were deployed to cover the use cases introduced in Section III.B. Prior to their deployment, both demonstrators were tested, first in laboratory, and then in near real conditions, in the facilities of the School of Engineering of the Universidad de Sevilla. This testing procedure was very important to train the TCS in order to build correct traffic scenarios, and to define and debug the heuristic and inference rules used in both layers of the DML. Fig. 6. A view of some of the SiUs used in the demonstrators: (a) the SiU close to the pedestrian crosswalk in the Suburban-Area Use Case, and (b) the SiU that informs about the presence of animals in the Rural-Area Use Case. Fig. 5. A view of the pole with the LC (including its IPU), the SeU, and the solar panels is shown. At the bottom, in a cabinet, the batteries and the supply unit. This demonstrator was deployed in Villafranca de Córdoba (Spain). Cameras Solar Panels LCU and IPU SeU Battery and Supply Unit a)a) (b) Fig. 6. A view of some of Fig. 6. A view of some of the SiUs used in the demonstrators: (a) the SiU close to the pedestrian crosswalk in the Suburban-Area Use Case, and (b) the SiU that informs about the presence of animals in the Rural-Area Use Case.SiUs used in the demonstrators: (a) the SiU close to the pedestrian crosswalk in the Suburban-Area Use Case, and (b) the SiU that informs about the presence of animals in the Rural-Area Use Case.