**1. Introduction**

In the event of severe vehicle collisions, the airbag deploys in 30–50 milliseconds and restrains the occupants providing a cushion effect [1]. An airbag offers optimum restraint effect when it deploys as designed for the situation; otherwise, there can be mortal injuries. A 20 ms late deployment can increase the risk of head injuries by 14% caused by the airbag and it also increases the risk of collision with the headrest [2]. Therefore, there is a need for control and optimization of the airbag deployment. Since the introductionof airbags, there have been many attempts to optimize the restraint effect and reduce the injuries for different crash situations by tuning various parameters such as airbag deployment time, early occupant coupling with the airbag, pressure dispersion direction and stagewise deployment [2–5]. Mercedes-Benz developed PRE-SAFE® Impulse Side. In this technology, the occupant is pushed forward during the potential crashes and engaged with the restraint system to reduce the kinetic energy difference between the occupant and the restraint system. The technology achieved a 35% reduction in upper rib displacement in a standard pole test [3]. Kim et al. designed a low-risk deployment airbag with a protective wrap. It disperses the airbag pressure in lateral directions and reduces the force on the occupants [4]. The self-adaptive vent (SAV) is a useful optimization technique for keeping the airbag inflated for a longer protection time. When the airbag is fully deployed, the tether tightens and closes the vent holes. The airbag remains inflated for a longer time and protects the occupants [5].

The mentioned state-of-the-art technologies do not consider occupant detection and classification (size and position of impact). The occupant contact data with the airbag and the bag pressure feedback are essential to make the airbag self-adaptive. Ultrasonic sensors, capacitive sensors, seat sensors, infrared cameras and computer vision systems

**Citation:** Shirur, N.; Birkner, C.; Henze, R.; Deserno, T.M. Tactile Occupant Detection Sensor for Automotive Airbag. *Energies* **2021**, *14*, 5288. https://doi.org/10.3390/ en14175288

Academic Editors: Guzek Marek, Rafał Jurecki and Wojciech Wach

Received: 27 July 2021 Accepted: 23 August 2021 Published: 26 August 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

can detect the occupants [6–14]. Izumi et al. developed an occupant detection system using a far-infrared camera, which provides the occupant's thermographic images. A vital shortcoming of the IR camera is its difficulty in tracking the occupant after the airbag deployment. Due to the overlap, the occupant or the airbag's contour cannot be obtained from the thermographic images. Dust particles and temperature changes make occupant tracking further complicated [6].

Computer vision systems have evolved. The occupant's presence and position can be detected using a stereo camera. If an occupant is present, the geometry and position of the occupant's head can be calculated further [11]. Farmer et al. developed a visionbased system to classify (adult or child) and track the occupant's motion. The airbag can be suppressed if an infant seat or an adult who is critically out-of-position is detected [12,13]. Further, adaptive airbag deployment (trigger time) decisions can be made by tracking the occupant's head's position and deciding in-position and out-of-position situations [10]. Airbag deployment power can cause mortal injuries if the occupant is in an improper posture. Hence, posture estimation is also crucial for the airbag deployment strategy [12,14].

The vision systems discussed in the literature support only the airbag deployment strategy, for example, trigger time. The airbag deployment strategy is purely based on a set of pre-determined inputs from the machine vision. Airbag-induced injury mitigation requires continuous occupant motion and contact feedback with the airbag during the ride-down phase. Hence, an alternative solution is required.

There is significant technological advancement in capacitive sensing methods to detect the occupant. Blackburn et al. developed an occupant sensing apparatus using a variable capacitor that gives the occupant's position relative to the cockpit. Airbag trigger time can be decided based on the sensed position, but occupant–airbag interaction is not addressed [7]. Kithil et al. developed a capacitance sensing array mounted at different locations in the cockpit. The sensors provide the occupant position measuring the dielectric change between the plates. However, the array gives only the position. It is hard to address when the occupant contacts the airbag and differentiate between the occupant and the airbag [8]. White et al. designed a capacitive sensor using a rendered portion of the airbag. There is conductive paint on the bag surface connected to a capacitive sensor circuit outside the airbag. The sensor gives the occupant's contact time with the airbag, but to control the airbag power, we need first contact, out-of-position, and the occupant's area [9].

#### **2. Problem Description and Research Contribution**

The research discussed in the literature focuses mainly on optimizing the airbag for a long-standing time and less impact force. On the other hand, occupant detection systems aid only airbag deployment decisions based on size, motion and posture. The sensing methodologies are successful until the first contact with the airbag. There is a lack of occupant sensing methods for in-crash and post-crash phases. Further, to continuously adapt the airbag power and mitigate the injuries, we need airbag–occupant interaction data, which is still an open research opportunity. The interaction between the occupant and the airbag can be addressed by answering the following questions.


In this work, we have developed novel single and matrix tactile sensors to detect the occupant and measure the contact parameters, especially contact time and area. The research work answers the stated questions and bridges the gap between pre-crash and in-crash occupant monitoring.

### **3. Method Overview**

This paper focuses on designing and testing capacitive tactile sensors, which answer the stated questions and provide more insight into the airbag–occupant interaction. The

sensors consist of a conductive woven fabric connected to a conducting thread. Two sensor variants discussed in this paper are a single sensor, which gives occupant contact time and contact area, and a matrix sensor, which additionally provides the position. Figure 1 shows the sensors' configuration. The sensors are integrated with the airbag and follow the airbag's shape during the deployment. In this work, an airbag with sensors is tested under low-velocity pendulum impact. Understanding the sensor's behavior for different external parameter changes and the airbag deployment phases is crucial. We studied various deployment phases like the textile unfolding, time-to-first-gas, bag inflation and the cable capacitance effects [15]. Time-to-first-gas and inflation do not influence the sensor, whereas the unfolding event and cable capacitance significantly affect the sensor. Additionally, the sensor was benchmarked with contact times from the camera [15]. This paper's main objectives are sensor calibration, occupant's contact detection, position estimation and contact area calculation. The contact area and position enable adult-child classification with out-of-position cases, which is crucial to modulate the airbag's pressure.

The results of this work have a significant impact on the vent hole control. The vent hole opening can be controlled based on the occupant's data to optimize the restraint effect and minimize the injuries [16].

**Figure 1.** Sensors' configuration: (**a**) single sensor; (**b**) matrix sensor.

#### **4. Capacitive Sensing**

The human body has 100 to 300 pF capacitance, which is used to sense the occupant contact with the airbag [17]. In this work, projected self-capacitance theory in loaded mode is applied [18]. A single sensing electrode is installed behind the airbag surface, which acts both as transmitter and receiver [18–24]. It is called active capacitive sensing due to the single electrode [18]. The sensor is connected to a resistor. The sensor's voltage is measured using an RC circuit. Suppose the occupant comes near the sensor, the capacitance increases due to the active coupling resulting in voltage drop [21]. The familiar examples of self-capacitance applications are a touch screen of a cell phone, sliders and wheels, control buttons on home appliances and the on-board infotainment of an automotive system [20,21]. Figure 2 shows the analogy between a capacitor and a capacitive sensor. A capacitor comprises two electrodes separated by a dielectric material. If a conducting electrode is removed, it assumes a virtual ground through the human body. This phenomenon is known as projected capacitance [18,19]. In our case, the sensor forms a conductive layer and the airbag textile forms the dielectric layer. A conducting thread is used for the connection between the sensor and the external circuit.

**Figure 2.** Sensor concept.

The sensor has a reference capacitance and voltage based on its geometry and dielectric values. Firstly, the reference value has to be measured to detect the touch. The voltage drop from the reference gives the touch. Figure 3 shows the capacitance variation as a function of occupant's distance from the sensor [22,25]. The capacitance increases when the occupant is in the sensor's proximity and observes a sharp gradient during the contact. There is a ΔC increase in capacitance, corresponding to ΔV voltage drop across the sensor.

**Figure 3.** Capacitance variation as a function of occupant's distance from the airbag.

Equation (1) gives the voltage across the sensor [19,20,24,25]. The voltages across the capacitor and the supply are *Vcap* and *Vs*, respectively. *t* is the time elapsed after the supply and *R* is the resistor's resistance. The permittivities of vacuum and dielectric material are *ε*<sup>0</sup> and *εr*, respectively. *A* denotes the plate area (contact surface area between the occupant and the sensor), and *d* is the separation distance (Figure 2).

$$V\_{cap} = V\_s \left(1 - e^{\left(\frac{-td}{Kda\_0a\_r}\right)}\right) \tag{1}$$

The sensor voltage behaviour without and with touch is illustrated in Figure 4. The green and blue curves show the voltage (*Vcap*) across the sensor without and with touch, respectively, for a supply voltage (*Vs*). In a transient state, there is a voltage drop of Δ*Vcap*. The time constant T can be varied by varying the resistance. Δ*Vcap* is used as a contact detection parameter. Based on this value, the contact area can be calculated. When there are multiple sensors, an appropriate threshold can be applied to determine the first contact point.

**Figure 4.** Capacitor charging curve with and without touch.

Equation (1) is implemented in Matlab® to simulate the sensor voltage behaviour as a contact area's function. In the theoretical simulation model, a 5 V source signal is given. The airbag fabric is 0.33 mm thick with *ε<sup>r</sup>* = 3.4. A 220 kΩ resistor is used in series with the capacitor. Figure 5a shows the simulated sensor voltage as a function of the contact area. 5 V is taken as a reference value. If there is a contact, for example, 0.01 m2, the voltage will be 2.5 V. The drop from 5 V to 2.5 V is the Δ*V* illustrated in Figure 5b. As the contact area increases, the voltage drop increases.

**Figure 5.** Sensor's voltage: (**a**) voltage–area relation; (**b**) voltage drop versus increase in contact area.

#### **5. Method**

#### *5.1. Sensor Hardware and Circuit*

The sensing surface consists of woven conductive fabric with a conductive thread (Figure 6). The fabric is a copper and nickel-plated nylon material that finds application in smart wearable technologies [26]. The thread is made of 30% stainless steel fibers [27].

The sensor dimensions are chosen to cover the full human face and partly the neck. The single sensor is a square with 200 mm sides and the matrix sensor has four individual sensors. Each sensor has a square shape with a 100 mm edge (Figure 1). A 20 mm gap is maintained between the sensors to avoid mutual capacitance and mutual touch.

Figure 7 shows the sensor's circuit diagram [21,23,28]. It is a simple RC circuit. The resistance has to be chosen based on the charging time. A smaller value requires an increased sampling rate; hence a 220 kΩ is selected. A 1 pF capacitor is connected to the ground to stabilize the sensor. The sensing surface is mounted on the airbag. The circuit is implemented using an Arduino development board as shown in Figure 6.

**Figure 7.** Sensing circuit.

#### *5.2. Algorithm for Sensor Voltage Change Measurement*

Digital input and analog output channels on the Arduino board are utilized to measure the voltage change. The channels are pulled down with internal resistors to avoid floating. A 5 V input signal is given for 136 microseconds. The voltage change across the sensor is measured from the analog channel. After the measurement, input and output channels are discharged and the next cycle is executed.
