Over the last decade, sensor technology has become ubiquitous and has attracted a lot of attention. Sensors have been deployed in many areas such as healthcare [
7,
8], agriculture [
9,
10], and forest [
11,
12], vehicle and marine [
13,
14] monitoring. In transportation, sensor technology supports the design and development of a wide range of applications for traffic control, safety, and entertainment. In recent years, sensors, and actuators such as tire pressure sensor and rear-view visibility systems have become mandatory (due to federal regulation in the United States [
15]) in the manufacturing of vehicles and the implementation of intelligent transportation systems, aimed at providing services to increase drivers’ and passengers’ satisfaction, improve road safety and reduce traffic congestion. Other sensors are optionally installed by manufacturers to monitor the performance and status of the vehicle, provide higher efficiency and assistance for drivers. Currently, the average number of sensors in a vehicle is around 60–100, but as vehicles become “smarter”, the number of sensors might reach as many as 200 sensors per vehicle [
16].
2.2. In Road Sensors
Strategic investment in transportation infrastructures is vital for a country’s growth and is the central core of a modern economy. Each year, governments worldwide spend a huge amount of money in the transportation sector. In the United States, the yearly investment is around 1.6 percent of the Gross Domestic Product (GDP) [
22] and Europe invested around 102 billion euros in 2014 with 52% spent in road infrastructures [
23]. Even though the automotive industry has invested a lot of money to increase safety, performance and comfort in vehicles using sensors within the vehicle; traffic data collection using mechanisms located along the roadside has become one of the main challenges for intelligent transportation systems. Sensor deployment within a transportation network provides drivers with new services such as smart parking (e.g., matching drivers with available parking spots) and reduced pricing according to congestion levels on the road. Sensors collect environmental data in real-time which is then processed and analyzed to improve transportation networks and make them resilient.
Sensors can be classified into two categories based on their location: intrusive and non-intrusive [
24]. Intrusive sensors are installed on pavement surfaces. They have high accuracy, but they also have high installation and maintenance costs. Basically, intrusive sensors (as shown in
Figure 2) can be classified into three groups: (1) passive magnetic sensors which are installed on roads and are connected either wired or wirelessly to processing units (2) pneumatic tube sensors placed across the road which transmit data to processing units through wired/wireless media, (3) inductive loops that are wire coils buried into roads and send data to processing units. This group of sensors is the most used in traffic control systems [
25].
The main advantage of road sensors is their technology maturity. They have been widely implemented and have high accuracy in detecting vehicles. However, the main disadvantages of road sensors are: high installation costs, traffic disruption during installation, maintenance, and repairs. One solution that has been implemented to address the aforementioned drawback is the introduction of wireless battery-powered sensor nodes which replace the intrusive sensors and are installed over the pavement. This technology represents a change in the transportation sensors which are expected to improve the quality, quantity, accuracy, and trustworthiness of the data collected from roads and avenues at a lower cost than current solutions [
26,
27,
28].
Non-intrusive sensors are installed at different places on the roads (other than over it) as shown in
Figure 3 and could detect a vehicle’s transit and other parameters such as vehicle speed, and lane coverage. However, they are expensive and may be affected by environmental conditions. Normally, non-intrusive sensors are used to develop applications that provide information on a selected location, such as queue detection at a traffic light, traffic conditions, weather conditions of the road and the pavement. Some sensors are mounted on a mast and are used to monitor a specific coverage area. Other sensors are mounted on bridges with a monitoring area directly below. Finally, some sensors are placed road side at ground level and use a beam that crosses the road and are mainly used for a single lane and with unidirectional flows because they are very susceptible to interferences from other objects.
Non-intrusive sensors provide many of the intrusive sensors’ functions with fewer difficulties. However, they are highly affected by climate conditions such as: snow, rain, and fog, among others. Accurate traffic data is of utmost importance to make informed decisions to improve traffic conditions. Non-intrusive sensors are more easily spotted by drivers, resulting in different and faster reactions such as: slowing down, and using the correct drive lane, among others after detecting those devices. The challenge is not just the installation of these sensors, but also reducing the drivers’ reactions times based on the collected data and provide them with a more precise view of the context and the reality of the road or avenue.
Currently, several sensors are used on roads.
Table 2 shows the two categories (intrusive and non-intrusive) of sensors that are currently used for keeping track of the number of vehicles, vehicle classification, or road conditions [
29] as well as some other practical uses.
Pneumatic road tube sensors use one or several tubes placed across traffic lanes allowing for number of vehicles counting and vehicle’s classification. When a vehicle’s tire passes over the tube, the sensor sends a burst of air pressure which produces an electrical signal. The electrical signal is transmitted to the processing unit.
The Inductive Loop Detector (ILD) sensor is one of the most common sensors in traffic management. It is used for collecting traffic flow, vehicle’s occupancy, length, and speed. It consists of a long wire coiled to form a loop which is installed into or under the surface of the road and measures the change in the electrical properties of the circuit when a vehicle passes over the sensor, producing an electrical current that is sent to the processing unit.
Magnetic sensors are used to detect vehicles when a change in the earth’s magnetic field is produced. Magnetic sensors are used to collect flow, occupancy, vehicle length and speed and are suitable for deployment on bridges.
Piezoelectric sensors detect vehicles passing over (at high speed ranges around 112 km/h) a sensor though a change in the sensor’s voltage and can monitor up to four lanes. Piezoelectric systems are commonly formed by piezoelectric sensors and ILDs sensors.
A Video Image Processor (VIP) system includes several video cameras, a computer for processing the images and a sophisticated algorithm-based software for interpreting the images and translating them into traffic data. Video cameras placed at the roadside collect and analyze video images from a traffic scene to determine the changes among successive frames using traffic parameters such as flow volume and occupancy. The main disadvantage of VIP systems is that they are susceptible to reduced performance caused by bad weather conditions.
Radar sensors transmit low-energy microwave radiation that is reflected by all objects within the detection zone. There are different types of radar sensor systems: (1) Doppler systems that use the frequency shift of the return to track the number of vehicles, and determine speed very accurately, (2) frequency-modulated continuous wave radar radiates continuous transmission power such as a simple continuous wave radar and is used to measure flow volume, speed, and presence. In general, radar sensors are very accurate and easy to install. They support multiple detection zones and can operate during the day or night. Their main disadvantage is high susceptibility to electromagnetic interferences.
Infrared sensors detect the energy generated by vehicles, road surfaces or other objects. Basically, sensors convert the reflected energy into electrical signals that are sent to the processing unit. Infrared sensors are divided into two categories: Passive Infrared (PIR) detects vehicles based on emission or reflection of infrared radiation and are used to collect data from flow volume, vehicle presence and occupancy. Active InfraRed (AIR) sensors use Light Emitting Diodes (LED) or laser diodes to measure the reflection time and collect data on flow volume, speed, classification, vehicle presence, and traffic density.
Ultrasonic sensors calculate the distance between two objects based on the elapsed time between a sound wave transmitted at frequencies between 25 and 50 KHz and reflected to the sensor by an object. The received energy is converted into electrical energy which is sent to the processing unit. Ultrasonic sensors are used to collect data about vehicle flow and the vehicle’s speed. The main disadvantage of this kind of sensors is its high sensitivity to environmental effects.
Acoustic array sensors are formed by a set of microphones that are used to detect an increase in sound energy, produced by an approaching vehicle passing through the coverage area of the sensor. Acoustic sensors are replacing magnetic induction loops to calculate traffic volume, occupancy, and average speed of vehicles.
Road surface condition sensors use laser and infrared technologies to read road conditions (temperature and grip) to improve traffic safety and execute road maintenance programs. However, this type of sensor requires periodic maintenance to maintain its performance level.
Radio-Frequency ID (RFID) sensors are used for: (1) automatically identifying running vehicles on roads and collecting their data, (2) for smart parking and for detecting vehicles to allocate space for parking.
Even though many sensors have been installed in roads and streets, the lack of a correct calibration and cluster integration makes the data collected unstable and hinders the development and evolution of ITS as projected and expected from transportation authorities, car makers, road users, and all ITS stakeholders. ITS are expected to use all kinds of integrated sensors to provide situation evaluation systems, and fast decision-making based on the data collected from integrated sensors to improve transportation conditions.
2.3. Discussion about Key Sensors
Although sensors are available and widespread, they are a small portion of the various types of equipment used in the automotive industry’s future planning: the self-driving vehicle. The redundancy and integration of sensors will improve the safety and performance of self-driving, automated or autonomous vehicles. Today’s vehicles are already equipped with radar and camera systems, redundant sensors, and software to control them. However, high-resolution and affordable LIDAR systems with ranges up to 300 m and higher are still at the pre-development stages, but it is foreseen that such components will become more advanced in the next few years.
Current camera systems are using CMOS image sensors and machine vision integration with multiple sensors (such as RADAR) for Advanced Driver Assistance Systems (ADAS) and partial autonomous driving. The main disadvantages of camera systems are that environmental conditions can cause problems in the detection of objects in non-illuminated and varying lighting conditions, and computer vision limitations for reliable detection. The challenge of camera systems is generating fast image acquisition and efficient image processing approaches for real-time analysis.
RADAR sensors utilize radio waves to measure the distance and are used with external controllers to modify the throttle to maintain a constant distance from an object. The benefit of RADAR is its low weight and a capability to operate in different conditions, but its main disadvantage is that it has a limited field of vision (small vehicles such as bicycles or motorcycles that are not moving in the center of the lane may not be detected).
LIDAR is a new system in the automotive sector used to measure distances to stationary as well as moving objects. LIDAR employs special procedures to provide three-dimensional images of the objects detected. The main disadvantages of LIDAR are their size, cost, and their limited capabilities in bad weather conditions (snow, fog, rain, and dust particles in the air) because LIDAR uses light spectrum waves. LIDAR is not able to detect a color, or contrast, and yields poor optical recognition. Finally, LIDAR systems today are only very rarely used on a large-scale production. Consequently, the potential of this technology is not yet fully explored because of cost and availability reasons. The challenges for LIDAR technology are to reduce the cost of deployment in all vehicles by reducing their size to enable easy integration in cars and higher aperture angle positions.
Companies such as Tesla are focusing on the development of new vehicles based on systems that contain only cameras and RADAR sensors whereas Google is using LIDAR as the preferred technology in the Waymo project [
30].
There is no unique solution for implementing assistance systems in vehicles. The success of the new era of vehicles is based on the integration of multi-sensor systems. Camera systems are being integrated with radar systems (located in the front and back of the vehicle to monitor traffic) to improve the precision of measuring of speed and distance and the detection of the outlines of obstacles and moving objects. Radar does not necessarily give the granularity provided by LIDAR, but Radar and LIDAR are promising technologies that complement each other well.