Implementation and integration of an experimental vehicle sensor setup for automated parking

Implementation and Integration of an Experimental Vehicle Sensor Setup for Automated Parking Ertüchtigung und Aufbau eines Versuchsträger-Sensor-Setups für automatisiertes Parken Bearbeiter: Alberto Linares | 2288833 Betreuer: Philipp Rosenberger, M.Sc. Alberto Linares Matrikelnummer: 2288833 Studiengang: Allgemeiner Maschinenbau Bachelor-Thesis Nr. 1318-18 Thema: Implementation and Integration of an Experimental Vehicle Sensor Setup for Automated Parking Eingereicht: September 17, 2018 Technische Universität Darmstadt Fachgebiet Fahrzeugtechnik Prof. Dr. rer. nat. Hermann Winner Otto-Berndt-Straße 2 64287 Darmstadt Erklärung zur Bachelor-Thesis Hiermit versichere ich, Alberto Linares, die vorliegende Bachelor-Thesis gemäß § 22 Abs. 7 APB der TU Darmstadt ohne Hilfe Dritter nur mit den angegebenen Quellen und Hilfsmitteln angefertigt zu haben. Alle Stellen, die aus Quellen entnommen wurden, sind als solche kenntlich gemacht worden. Diese Arbeit hat in gleicher oder ähnlicher Form noch keiner Prüfungsbehörde vorgelegen. Mir ist bekannt, dass im Falle eines Plagiats (§38 Abs.2 APB) ein Täuschungsversuch vorliegt, der dazu führt, dass die Arbeit mit 5,0 bewertet und damit ein Prüfungsversuch verbraucht wird. Abschlussarbeiten dürfen nur einmal wiederholt werden. Bei der abgegebenen Thesis stimmen die schriftliche und die zur Archivierung eingereichte elektronische Fassung gemäß § 23 Abs. 7 APB überein. Darmstadt, den 4.2.2016 (Alberto Linares) Abstract This thesis has been realized for the automotive engineering department of the TU-Darmstadt University. This department is currently carrying out research in the field of sensor modeling to support autonomous driving. During this thesis the main objective is the rear sensors Bosch parkpilot URF7 implementation and integration, using the methodology described by the thesis1. Besides, part of the results obtained have been used for the modeling of ultrasonic sensors for valet parking use cases. The thesis is divided in different stages. First of all, a documented bibliographic search will be carried out on the ultrasonic sensors general operation and the different most important technical specifications of an ultrasonic sensor in the field of the automotive industry as well. Among them are: the field of view, detectable objects distance ranges and accuracy of the measurement. Once the necessary information for the thesis has been obtained, the Bosch Parkpilot URF7 has been used and the different tests previously described in the methodology research have been elaborated. Besides, imparting of the initial methodology extra tests have been made based on the calculation of the distance between sensor and object in real time. It should be noted that all these tests have been run without the sensor setup installation in the car. In contrast, the position and orientation of the sensors has been fixed as in the institutes test vehicle, Honda Accord, thanks to a movable platform. Once the system has been installed and powered with an external power supply, the basic functionality as an user level has been checked, in other words, the system calibration and the operating mode. To get a better understanding of the operation, the board that controls the Bosch sensor setup has been examined. The actual functioning of the existing hardware could not be found successfully due to the lack of information from the manufacturers. However, it is known the existence of a communication between sensors and the control unit, so that, with the realization of different tests and observing its signal with a DAQ, the signals can be read an processed. Once the operation of this Bus has been checked, the different experiments are performed to determine some technical specifications of the sensors: the field of vision (both of a sensor and of the entire sensor setup), the distance range in which a object can be detected, the operation of the cross echo and the accuracy in the measurement. Finally, with an Arduino UNO the signal of the communication bus has been processed to obtain the distance to an object as accurate as possible in real time. In addition, the sensitivity of the results obtained with the Arduino will also be determined. 1Fu, J. et al.: Setup for ultrasonic models validation (2017) Contents 1 Introduction 1.1 Motivation ................................................. 1.2 Methodology................................................ 1.3 ConcretionofAssignment........................................ 2 Ultrasonic Sensors 2.1 ConversionPrinciples........................................... 2.1.1 PiezoelectricEffect ....................................... 2.1.2 PiezoelectricMaterial...................................... 2.2 UltrasonicTransducer .......................................... 2.2.1 ActiveElement.......................................... 2.2.2 BackingandWearPlate .................................... 2.2.3 EquivalentCircuit........................................ 2.3 DistanceMeasurements ......................................... 2.3.1 DistancesBetweenPulses ................................... 2.3.2 Object Localization and Trilateration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Bosch Parkpilot URF7 3.1 TechnicalSpecifications ......................................... 3.2 SystemCalibration............................................ 3.3 BoschParkpilotURF7Operation................................... 4 Bosch Parkpilot URF7 Tests 4.1 HardwareandSoftware ......................................... 4.1.1 PowerSupply........................................... 4.1.2 National Instrument DAQ and Labview . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1.3 Arduino .............................................. 4.1.4 StaticTestsSetup ........................................ 4.2 COMMBusOperation.......................................... 4.3 DistanceMeasurementTest....................................... 4.4 Maximum and Minimum Range Distance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.5 Sensor Field of View Determination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.5.1 CrossEchoOperation...................................... 4.5.2 Whole Setup Field Of View Determination . . . . . . . . . . . . . . . . . . . . . . . . . 4.6 RealTimeDistanceMeasurement................................... 4.6.1 HardwarePreparation...................................... 4.6.2 SoftwareOperation ....................................... 4.6.3 SensitivityMeasurement .................................... 5 Conclusions 6 annexes References Version: September 17, 2018 List of Figures 1 Emitter and receiver Bosch ultrasonic sensor. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Emitter and receiver HC-SR04 sensor. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Vertical/Horizontal ultrasonic sensor field of view. . . . . . . . . . . . . . . . . . . . . . . . . . 4 Piezoelectriceffect ............................................ 5 Inversepiezoelectriceffect........................................ 6 Titanate of lead zirconate in crystalline perovskite structure. Above and below the Curie temperature ................................................ 7 Different ultrasonic transducer parts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 Transducerequivalentcircuit...................................... 9 Ultrasound wave and its envelope wave. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 Trilateration obstacle distance measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 BoschParkpilotURF7. ......................................... 12 LEDs distribution and sensor wiring harness . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 BoschParkpilotURF7operation ................................... 14 FinalSetupfortheexperiments .................................... 15 Experimentsblockdiagram....................................... 16 Voltcracftpowersupply ......................................... 17 NI hardware and Labview softwae implemented . . . . . . . . . . . . . . . . . . . . . . . . . . 18 ArduinoUNORev3board ....................................... 19 Previousstatictestssetup........................................ 20 Layoutforthesetupbar. ........................................ 21 Newsetupbar............................................... 22 Fixingpiecesforsensors......................................... 23 COMMBussignaldifferentpulses .................................. 24 COMM Bus signal no object detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 COMM Bus signal one object detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 Two objects detected in the same period . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 Walltestsetup............................................... 28 Object detection probability in ten periods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 Field of View experiment procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 Sensor position to determine both FoV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 Point objects used for the experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 Echo behaviour depending on the shape . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 Bosch sensor horizontal field of view . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 Boschsensorverticalfieldofview................................... 35 Cross-EchoSetup.............................................. 36 COMM Bus signal cross echo for different scenarios. . . . . . . . . . . . . . . . . . . . . . . . 37 WholesystemFoVmetalstick..................................... 38 WholesystemFoVplasticstick .................................... 39 Voltagedividerschematic........................................ 40 Arduinoconnexions............................................ 41 Example of a large object situation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 Samples in a short distance and N=20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 Samples in a medium distance and N=20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Version: September 17, 2018 44 Samples in a large distance and N=20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 Samples in a short distance and N=50 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 Samples in a short distance and N=50 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 Samples in a short distance and N=50 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 Samples in a short distance and N=100 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 Samples in a medium distance and N=100 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 Samples in a large distance and N=100 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 Samples in a short distance and N=200 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 Samples in a medium distance and N=200 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 Samples in a large distance and N=200 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 Samples in a short distance and N=400 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 Samples in a medium distance and N=400. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 Samples in a large distance and N=400 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Version: September 17, 2018 List of Tables 1 BoschParkpilotURF7techspecs.................................... 2 Arduino UNO Rev3 technical specifications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 FoVBoschsensormeasurements. ................................... 4 Metal stick whole FoV measurements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Plastic stick whole FoV measurements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 Measured value for a short actual distance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 Measured value for a medium actual distance. . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 Measured value for a large actual distance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Short distance sensitivity experiment results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 Medium distance sensitivity experiment results. . . . . . . . . . . . . . . . . . . . . . . . . . . 11 Long distance sensitivity experiment results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . Version: September 17, 2018 1 Introduction Nowadays there is a high interest to make cars as autonomous as possible and new technologies are helping to achieve it. The investments for developing fully autonomous cars are increasing day by day and did not reach their peak yet. With the responsibility of driving human beings, these cars have to be tested absolutely carefully. However, due to the advanced technology, testing is highly costly and time-consuming, limiting the number of testable scenarios. Talking about parking scenarios, there are a lot of human drivers out there that can not quite handle the parallel parking job, mainly because they lack a perspective view on the object’s surrounding them. It especially occurs in big cities with lots of cars and tight spaces for parking causing traffic tie-ups, vehicle damage. To help human drivers in such situations, Advanced Driver Assistance Systems (ADAS) have been developed. Systems which help you to drive, in this case, help you to park. ADAS warn you of the object proximity while you are parking with an acoustic signal. Others are more sophisticated and use cameras showing the human driver a real time picture of the car’s surrounding. Latest tech systems are even able to park the car autonomously. 1.1 Motivation The motivation of this thesis is presented as a contribution for the before mentioned problems, since together with a master thesis2that is running in parallel, the final objective is to validate a ultrasonic sensor model in valet parking use cases. Besides, it is colaborated for the ENABLE-S3 project in the development of simulation environments with the purpose of implementing an autonomous parking system. Therefore, all those problems that the human driver has when is parking will disappear. Besides, being able to recreate with the simulations environment infinite situations in a more efficient, cheap and fast way with the purpose of improving and checking all these systems, seeing that real environment tests are too expensive3. 1.2 Methodology This thesis has had a duration of 5 months in which differents methods has been followed to arrive at an objective with final results and conclusions. Note that for the collection of these results has been needed a power supply, which simulated the reverse gear feeding thus the setup. In addition, a DAQ connected to a computer, which had a Labview script implemented for the acquisition of data from the sensor setup, was needed. Remark that the chosen software was Labview due to its easy implementation with the DAQ and because it has implemented a graphical interficie that shows the information in a very visual way. To know the operation of the communication between sensor-ECU a deductive method was applied, since from the basic general knowledge of the ultrasound sensors the operation of this was deduced. Observing the signal while a object was placed at different distances and positions or even removing the object. Then, to determine the different technical specifications, a method of experimentation has been applied, in other words, some differents experiments were carried out to determine them. Finally, a method of collecting data and sampling these in real time will be realized with a microprocessor. In this case an 2Fu, J.: Ultrasonic sensor model (2018) 3Kishonti, L.: Real and simulated tests (2017) Version: September 17, 2018 countermass is placed on the back face, whose fundamental objective is to absorb the mechanical energy in that direction and stop the oscillation of the ceramic, originating a transducer with higher resolution. The Wear plate on its part has two functions, protecting the active element and ensuring greater energy transfer, the latter is achieved by manufacturing it from a material with an intermediate acoustic impedance between the active element and the material on which it is expected to use the transducer17. 2.2.3 Equivalent Circuit A piezoceramic ultrasonic transducer can be shown close to its resonance frequency by an electrical equivalent circuit consisting of a resonant circuit in series with a parallel capacitance called C0, which is the disk capacitance of the piezoceramic (Figure 8). The value of C0in the adhesive bonded state of the ceramic is considerably lower than before bonding in normal circumstances. Due to C0has to show positive temperature dependency, this effect must be compensated by a parallel capacitance with a negative temperature dependency. In this way the resonance frequency of the electrical circuit can be kept stable with regard to the temperature. The resonance frequency is determined by f s =1 2πpLsCs where Ls and Cs are mechanical properties of the diaphragm. Figure 8: Transducer equivalent circuit18. 2.3 Distance Measurements There are different ways of generating and receiving ultrasonic waves for distance measurement applications. Commonly, continuous waves or waves in the form of a pulse can be used. There are essentially two continuous wave methods to calculate distances19: 1.) Based on the measurement of the phase difference between the transmitted signal and the received signal of an amplitude-modulated wave. 2.) Based on the measurement of the frequency difference between the transmitted signal and the the signal received from a frequency modulated wave. However, most of the distance measurement applications by ultrasound are based on the time estimation that there is between the emission of a short train of pulses of ultrasonic waves, and its reception after having been reflected by some object in the environment . This period of time is commonly referred to as flight time (TOF). 17 Rubio, C.: Ultrasonic transducer for automated sytems (2018) 18 Winner, H.: Handbook of Driver Assistance Systems (2014) 19 Navarro: Ultrasonic distance measurement (2004) Version: September 17, 2018 2.3.1 Distances Between Pulses Sensor generates an ultrasonic pulse which is transmitted through the medium (typically air) until it is reflected by some reflecting surface. By measuring the time between transmission and reception of the echo, the distance to the reflector can be estimated indirectly by D=v·tf 2, where vrepresents the speed of sound in the transmission medium and the flight time20: The accuracy in the measurement of distances using this technique depends on the knowledge of v and the correct estimate of t f . The speed of sound in the air shows an almost linear dependence on temperature, which can be easily determined. Sound speed in the air is 343.2 m/s at 20ºC. Then the critical point in the measurement of distances using this technique is the determination of the time of flight. The most common way to determine flight time is by the threshold method, in which the arrival time is calculated when the echo received for the first time passes a certain level of given amplitude. Figure 9: Ultrasound wave and its envelope wave21. 2.3.2 Object Localization and Trilateration Car bumper is composed by 4-6 sensors both the front and the rear one. When it is time to measure the distance from an object to the bumper, it is assumed that the objects are divided in two big groups. On one hand, there are extended obstacles, these can be, for example, a wall or a vehicle. If this is the case, then the shortest measured distance also corresponds to the actual distance. Therefore, the formula that is mentioned in the section 2.3.1 is the one that has to be used. On the other hand, the second case is when it is a unique object. Since the distance that each sensor calculates is not the nearest from the bumper, Pythagoras theorem (1) is applied to established the distance from the object to the bumper22. D= √DE12− (d2+DE2+DE22)2 4d2(1) 20 Carullo, A.; Parvis, M.: Ultrasonic sensor distance measurement (2001) 21 Navarro: Ultrasonic distance measurement (2004) 22 Winner, H.: Handbook of Driver Assistance Systems (2014) 23 Winner, H.: Handbook of Driver Assistance Systems (2014) Version: September 17, 2018 Figure 10: Trilateration obstacle distance measurement23. 3 Bosch Parkpilot URF7 Bosch Parkpilot URF7 setup is a parking rear aid system including four ultrasonic sensor and an ECU (Electronic Control Unit). The setup warns the driver by using an acoustic signal with a buzzer and indicative LED lights, that also indicate the range of distances. In addition, it has fixing accessories to fix the sensors in the bumper and a tutorial CD. Figure 11: Bosch Parkpilot URF7. The four sensors of the setup are the same. Each sensor has a cable that consists of three wires which are equivalent to the three pins that each sensor has. The power pin (VCC), the ground pin (GND), and the communication pin (COMM). The COMM pin is used to establish a communication between the different sensors and the ECU. Moreover, the cables of each sensor are grouped into one and connected to the ECU. On the other hand, the buzzer and the LEDs cables are directly connected to the ECU. Also the setup has a diagnostic wire to know where the error comes from in case the system fails. Finally, the Bosch Parkpilot has the GND wire and the power wire, which is connected to the reverse gear. Version: September 17, 2018 3.1 Technical Specifications The Bosch system technical specifications24 are in the Table 1: Table 1: Bosch Parkpilot URF7 tech specs. Service Voltage from 9V until 16V Maximum Current Consumption 200mA Service Temperature -40°until +85° Maximum Distance 1500 mm Minimum Distance 300 mm Beam Angle No Available 3.2 System Calibration Once the setup is mounted in a car, a calibration of the system has to be done since each car has a different bumper. The following steps are the ones that are needed to calibrate it. 1.) Place the car in front of a wall in a distance of approximately two meters. 2.) Turn on the car and set the reverse gear. Now, LED A + D are turned on (start-up mode). 3.) Drive slowly backwards in a straight line towards the wall, until the following indication can be seen: LED A + D continues to shine and, additionally, first both yellow B + C LEDs flash and then shine permanently. 4.) When the four LEDs A + B + C + D are permanently lit, stop the vehicle, set the parking brake, pull out the reverse gear and turn off the car. 5.) Turn on engine. 6.) Set the reverse gear. Now the automatic calibration procedure starts; which is indicated by the LED flashing in pairs for 45 seconds. The successful calibration is shown by the following designated sequence of LEDs: LED B + C + D + E light up. 7.) Remove the reverse gear. 8.) Move the vehicle away from the wall 2 meters approximately. 9.) Turn off engine. 10.) Cut the cable (BK = black) on the sensor wiring harness, see the Figure 12. 11.) Finally, check if the system is working properly. Turn on engine. 12.) Set the reverse gear. All LEDs flash briefly. The tone of service readiness is heard. The system is now ready for use. 24 Bosch Operating Intructions. Version: September 17, 2018 Figure 12: LEDs distribution and sensor wiring harness25. 3.3 Bosch Parkpilot URF7 Operation The operation of parkaid Bosch system once it is installed in the car is not complex. As the car approaches an object, the LEDs start turning on as a warning for the driver. Also, at distance of less than 700 mm between the car and the object an acoustic signal appears, which depending on the distance, will change the frequency. Distance is shown in steps of 300 mm to 100 mm thus it is not a very accurate system. The exact operation of the system is shown in Figure 13. Figure 13: Bosch Parkpilot URF7 operation26. 25 Bosch Operating Intructions. 26 Bosch Operating Intructions. Version: September 17, 2018 4 Bosch Parkpilot URF7 Tests During the thesis different tests have been run. The data collected from these tests will be used for a parking aid simulation environment implementation. Therefore, it will be as realistic as possible. The tests can be divided into two groups: the parkaid Bosch setup technical specifications and parkaid Bosch setup data processing and data collection. Note that, for the different tests, the four sensors have been placed in the same disposal as in the workshop test car, Honda Accord. For the realization of these tests, different hardware and software instruments were needed. The first group of tests are carried out to have a better accuracy of the Bosch setup technical specification since the one that is described by the parkaid Bosch setup itself in section 3.3 is not accurate enough. In addition, the results obtained have an effect on the environment simulation because the values should be implemented with the highest possible accuracy. First, the COMM Bus operation between the ECU and a sensor is studied. Once the COMM Bus operation is known, looking at its signal, the distance between object and sensor can be determinated. In addition, the maximum object detect distance, minimum object detect distance and field of view (FoV) have been determined. The setup to perform these experiments is shown in Figure 14. Figure 14: Final Setup for the experiments. On the other hand the second test group consists of collecting data from each sensor COMM Bus with an Arduino. This information is processed with a script that has been developed to obtain the distance from the sensor to an object in real time. Besides, another test will be done to calculate the sensitivity of the calculated distance and thus be able to add this to the environment simulation. The setup for these experiments is shown in Figure 15. Figure 15: Experiments block diagram. Version: September 17, 2018 4.1 Hardware and Software 4.1.1 Power Supply Rear park aid Bosch setup is supplied when the reverse gear is set. As the Bosch setup is outside the car, a power supply is needed. It provides a voltage of 13.8V. (Figure 16). Figure 16: Voltcracft power supply. 4.1.2 National Instrument DAQ and Labview Data collection is the process of detecting electrical or physical phenomena such as voltage, current, temperature, pressure or sound with a computer. A data acquisition system consists of sensors, data acquisition device and a computer with programmable software27. In the following tests the data collection of the electrical signal is taken with the multifunction DAQ NI USB6363 X-Series. About software, Labview has been used due to its easy synchronization with the DAQ. In addition, Labview allows to show the collected data in a more visual way thanks to its graphic interface. Therefore, a computer with Labview is needed. The software developed in Labview for the data sample is shown in Figure 17. Figure 17: NI hardware and Labview softwae implemented. 27 Data acquisition. Version: September 17, 2018 4.1.3 Arduino Arduino is an open source electronic platform based on hardware and software. Arduino will be the microcontroller chosen for the tests since it has the following advantages in comparison to others: 1.) The university owns an Arduino. 2.) As an open source platform, there are many resources available on Internet. 3.) Compared to other one development boards, Arduino and related products are relatively cheap but also of excellent quality. 4.) Instructions in the Arduino software language are not complex, with basic programming knowledge one can apply Arduino quickly. 5.) The program code is loaded directly on the Arduino board through a USB cable. 6.) Its technical specifications are powerful enough for the application to be carried out. 7.) Multi platform. The Arduino programming environment is executable in Windows, Macintosh OSX and Linux28. The Arduino board chosen is the Arduino Uno Rev3. This consists mainly of digital and analog inputs and outputs, microcontroller and USB interface. Figure 18: Arduino UNO Rev3 board29. Arduino UNO Rev3 technical specifications are shown in the Table 2. Note that Arduino can be powered through USB or its pins. 28 Arduino IDE. 29 Arduino UNO Rev3. Version: September 17, 2018 Table 2: Arduino UNO Rev3 technical specifications30. microcontroller Atmega328P Operating Voltage 5V Input Voltage (recommended) 7-12V Input Voltage (limit) 6-20V Digital I/O Pins 14 Analog Input Pins 6 DC Current per I/O Pin 20mA DC Current for 3.3V Pin 50mA Flash Memory 32 KB SRAM 2 KB EEPROM 1 KB Clock Speed 16 Mhz Arduino IDE is the software that is used by default for programming with Arduino. This software uses a simplified C and C++ language, since it has different libraries included. It is based on the Processing environment as well as a programming language based on Wiring. The Arduino IDE comes with a code editor and integrates gcc as a compiler31. 4.1.4 Static Tests Setup A setup previously assembled by other students is used to fix the sensors position during the experiments and thus simulate the bumper of a car (Figure 19). The front bar of this setup has been modified in such a way that the sensors are located in the same position as institute’s Honda Accord test car. Therefore, distances have been taken between the four rear sensors of the car and a new front bar has been designed. Note that the base can be moved small distances and the front bar can take different angles thanks to a rotating mechanism. Figure 19: Previous static tests setup32. 31 Arduino IDE. 32 Fu, J. et al.: Setup for ultrasonic models validation (2017) Version: September 17, 2018 The distance between the central sensors of the institute car is 50 cm. Whereas the distance between the central sensor and the side sensor is 41 cm. Note that, the side sensors are 7 cm higher and 5 cm behind than the middle ones. Once the distances are known a new bar is designed keeping the same distances between sensors(Figure 20). Figure 20: Layout for the setup bar. It has been decided to apply an angle between both bars to prevent the sensors from detecting their own bar all the time. In case both were located with an angle of 90°with respect to them, sensors will detect all the time these ones. Figure 21: New setup bar. Finally, thanks to the shape of the metal bar with the pieces that are shown in Figure 22, screws and plastic ties sensors can be fixed in the bar. Figure 22: Fixing pieces for sensors33. 33 Fu, J. et al.: Setup for ultrasonic models validation (2017) Version: September 17, 2018 Figure 34: Bosch sensor vertical field of view. 4.5.1 Cross Echo Operation The cross echo occurs when a sensor receives the signal bounced by an object and this signal came from another sensor. To detect the cross echo in the COMM Bus, first an object is placed in front of one of the setup sensors, in such a way that this object is not detected by any of the other sensors in the setup. The sensor’s 1,3 and 4 signal COMM Bus is watched with Labview Figure 24. Therefore, the sensors are not detecting any object, while the signal of sensor 2 shows that it detects an object (Figure 25). This first configuration is done just to compare the COMM bus signal when it is detecting a cross-echo. Figure 35: Cross-Echo Setup. To facilitate the detection of a cross echo, an object is placed between sensor 1 and sensor 2 (Figure 35). While watching the COMM Bus signal of sensor 2 it can be seen that a period is formed by five pulses that come from the ECU instead of three. Version: September 17, 2018 Figure 36: COMM Bus signal cross echo for different scenarios. In the left picture of Figure 36, it is presented the COMM Bus signal of a sensor in a cross echo scenario where that sensor is receiving a cross echo from another sensor and also its own echo. It is known because after the last two large pulses it can be viewed two short pulses, the sent and its own echo whereas in the two middle large pulses it can be seen just one short pulse which is the cross echo. On the contrary on the right picture, the performing sensor is just receiving the cross echo. 4.5.2 Whole Setup Field Of View Determination Even though the field of view of a sensor has been determined, replicating this for each of the four sensor can not be applied for the whole system field of view. Since for intermediate zones between sensors, a sensor is able to receive the cross-echo and not its own echo. Which means that for only one sensor that object would be outside of its field of view, while for the whole setup that object would be inside the field of view. Therefore, the same experiment has been repeated as in section 4.5, but with the four sensors in operation. The four communication buses were observed at the same time to know if any of the sensors was receiving its own echo or cross-echo and thus determining the limit of the field of view. This experiment has been done with the metal stick with square shape and the plastic stick with round shape. The results obtained are the following Figure 37 and Figure 38 . Figure 37: Whole system FoV metal stick. Version: September 17, 2018 Figure 38: Whole system FoV plastic stick. The measurements have been taken for several values of d, in the case of the metal stick, d has a range of 0-2500 mm in steps of 100 mm-200 mm, while for the plastic stick, d has a range from 0-1600 mm in steps of 100 mm-200 mm. The plastic stick range is smaller because the setup can not detect it a distance further than 1600 mm. These measurements have been taken for the three different zones that are pointed out in both Figure 37 and Figure 38. For zone 1, the reference plane for the measurements is the one of sensor 1, while for zone 2 and 3 the reference plane is the one of sensor 2. All the sensors Bus COMM signal have been watched at the same time since in zones between sensors, it would be possible that a sensor can receive a cross echo and not its direct echo. The right half plane has been replicated, assuming the ideality of the sensors. Table 4: Metal stick whole FoV measurements. Zone 1 Zone 2 Zone 3 X (cm) d (cm) X (cm) d (cm) X (cm) d (cm) -44 15 -13 262 5 272 -60 30 -37 236 25 285 -79 47 -46 221 45 270 -88 66 -54 200 -92 82 -74 175 -96 100 -110 163 -114 120 -139 145 -96 133 -82 152 -55 170 -40 180 -21 200 -10 215 Version: September 17, 2018 Table 5: Plastic stick whole FoV measurements. Zone 1 Zone 2 Zone 3 X (cm) d (cm) X (cm) d (cm) X (cm) d (cm) -24.5 10 -2 170 5 272 -35 30 -27 150 45 285 -38 48 -40 130 25 270 -50 66 -78 120 -29 87 -105 101 -16 110 -2 120 4.6 Real Time Distance Measurement 4.6.1 Hardware Preparation Due to Arduino Uno Rev3 is not able to read voltages higher than 5V through its analog pins, and the COMM Bus signal assumes a range of values from 8.5V to 1V. For that reason, a voltage divider is used to make sure that the input of analog Arduino pin is not higer than 5V. In this way, it is possible to reduce the voltage by a factor that will depend on the resistances chosen for the formula (4)37. V0=R2 R2+R1 Vin (4) Figure 39: Voltage divider schematic. It has been decided to choose R1 = 8.2KΩand R2 = 10KΩas the values of the voltage divider resistors. These resistor values are chosen since for the maximum value of the input signal, in this case 8V, output voltage would be close to 5V which is what is interesting because with other values of resistances the range of Vout could be smaller being thus more difficult the detection of the different pulses. Note that, others couple of resistor would fix here as long as these obey the equation. Vin corresponds to the COMM Bus signal of each sensor and Vout corresponds to the same value as Vin but reduced by a factor 37 Krzentz, S. V.: Voltage divider (1998) Version: September 17, 2018 of 0.55. Therefore, the new range of values is 4.7V-0.55V. Now, Arduino is already able to read the input signal of its analog pins. The following figure shows the Arduino schematic (Figure 40). Figure 40: Arduino connexions. 4.6.2 Software Operation The software developed calculates the distance between each of the setup sensors to an object in real time, exactly showing the distance each period Tof the COMM signal. Remark that each period Tlasts approximately 250 ms. Arduino reads the COMM Bus signal from each of the sensors and processes it. It means that Arduino detects each of the different pulses, both from the ECU and from the sensor and after a sensor activation by the ECU, if Arduino detects two pulses coming from the sensor, the sent and received, it will calculate the distance in time between them by getting the timestamps of both pulses with the micros() function. Followed by applying the formula of section 2.3.1, Arduino returns the distance value in length units, exactly in centimeters. The micros() function returns the time value from when arduino is turned on until the instant in which the function is called. Besides, the software detects the cross echo but it does not calculate the distance between the sensor that has received the cross echo and the object. Note that the software measurements range goes from 19 cm to 3 m. Below, there are a tables comparing the actual distance with the measured distance when the object is placed in a short, medium and long distance from the setup. Version: September 17, 2018 Table 6: Measured value for a short actual distance. Real Distance (cm) Measured Distance (cm) 45 45.36 45 45.50 45 42.84 45 42.70 45 45.50 45 42.84 45 45.50 45 45.64 45 45.50 45 42.84 Table 7: Measured value for a medium actual distance. Real Distance (cm) Measured Distance (cm) 125 125.44 125 125.58 125 125.44 125 126 125 122.78 125 125.58 125 125.44 125 128.38 125 122.78 125 125.72 Version: September 17, 2018 Table 8: Measured value for a large actual distance. Real Distance (cm) Measured Distance (cm) 235 235.20 235 234.78 235 235.62 235 232.12 235 232.68 235 232.12 235 232.68 235 234.78 235 232.68 235 234.71 In the tables it is detected that for an actual distance X, the measured distance changes in a range of X+-3 cm. Besides, if the actual distance is changed just 1 cm the measured distance still being the same or jump a 3 cm step. It means that the systems now has a sensibility of 3 cm. With the purpose of fix it a diagnosis has been done and have been concluded that there are three possible cases where these errors may come from: 1.) Arduino is not enough fast to detect the change voltage signal, so that Arduino detects the sent pulse with a delay. The equation 1 f=Tis used to know how much Arduino needs to take a sample. Arduino works with a f=16Mhz. It means that Arduino takes a sample from the signal each 6, 25·10−8s. If the error of the distance measured is 3 cm, maybe Arduino has a delay detecting the received pulse. Applying the equation of section 2.3.1 is known that this 3 cm in time are 0.1765 ms (where v=340 m/s). Therefore, if Arduino takes a sample every 6, 25 ·10−8s, it will take more than one sample in 0.1765 ms. For that reason, Arduino is enough fast and the measurement problem does not come from it38. 2.) The pulse that corresponds to the direct echo hides information in its width. The duration of the pulse is measured for different types of objects but the same width result is always obtained, 3.9 ms. Accordingly with the result obtained,the measurement problem does not come from it either. 3.) The direct echo received by the sensor comes from different points where the signal has bounced. Depends on the signal bounced, the distance change its the value. The distance is calculated from more than one period Twithout moving the object position and the same distance is obtained always. With the aim of solving this error in the measurements, it is calculated the arithmetic mean of the different values in order to stabilize the final result (5). Consequently, now a final measurement is acquired depending on the selected mean whereas previously a final measurement was obtained for each period T of the COMM signal. The number of period Tper final result is equal to N39. Mean(X) =− x=∑N i=1Xi N(5) 38 Frequency and Period Time relation. 39 Arithmetic Mean Formula. Version: September 17, 2018 Nhas value of 400, 200, 100, 50 and 20 assigned, for long to short distances. In the section 6, results that have been obtained for each Nare shown. Note that for each of both Nvalue and kind of distance(long, medium or short) has been taken ten samples. In the results shown, it is observed that the distance from the sensor to the object is not a variable to taken into account. Also, the final measurements are stabilized in all cases, but logically, a greater stabilization can be seen for the N values of 400 and 200. Finally, the Nvalue chosen for the distance calculation is N=20. Since for a greater N, too much time is needed by Arduino to show the final measurement and this is not of interest for parking system. Once the Nvalue is chosen, the software is implemented in order to not calculate the distance between an object to a sensor, but the one between object and bumper. For this implementation, the objects are divided into two groups, large and small. It is assumed that an object is cataloged as large when 3 or more sensors are detecting it, whereas when only two sensors detect it, it is classified as small. When a large object is detected, the minimum distance between sensor and object will be shown, since it will be the same distance between object and bumper Figure 41. While when a small object is detected, the trigonometric relationship of section 2.3.2 will be applied to calculate the distance between the object and the bumper. In case only one sensor detects an object, the distance shown is the one that goes from the sensor to the object. Figure 41: Example of a large object situation. 4.6.3 Sensitivity Measurement This experiment consists of determining the sensitivity of the setup. Note that, with setting the sensitivy of the system you also set the accuracy accordingly and it will be implemented in the environment simulation. As software initial condicions for this experiment it does not contain the last part implemented about the distance from the bumper to an object. Besides, the Nvalue set is 400, since now the delay in obtaining the final measurement does not matter and it is useful to have a result as stable as possible. The procedure consists of placing an object at a Xdistance and take the result measured by Arduino. Followed by moving the setup 2 mm away from it, this is possible thanks to the rotating mechanism that the moveable platform of the setup has. Once the 2 mm are moved, the result are recalculated by Arduino. The process is repeated in steps of 2 mm until the measurement obtained by Arduino has changed to the next measurement step. Depends on this step, the sensitivity of the system based on the actual measurement is determinated. Finally it is viewed that when is changed 1 cm from the actual distance, the next measured step changes 1 cm as well. Version: September 17, 2018 This process has been repeated for long, medium and short distances and the results obtained are in the following Table 9, Table 10 and Table 11: Table 9: Short distance sensitivity experiment results. Actual Distance (cm) Measured Distance (cm) 44.8 43.61 45 43.63 45.2 43.93 45.4 44.07 45.6 44.73 45.8 44.89 46 45.43 Table 10: Medium distance sensitivity experiment results. Actual Distance (cm) Measured Distance (cm) 112.8 111.95 113 112.17 113.2 112.31 113.4 112.45 113.6 112.72 113.8 112.87 114 113.11 Table 11: Long distance sensitivity experiment results. Actual Distance (cm) Measured Distance (cm) 224.8 223.92 225 224.36 225.2 224.65 225.4 224.70 225.6 224.96 225.8 225.26 226 225.28 Due to the volatility for units smaller than the centimeters, it is decided to establish 1 cm as the sensitivity of the system. Note that the results obtained will be always one centimeter less than the actual distance. Version: September 17, 2018 5 Conclusions Once the methodology was followed and the different experiments were elaborated, the results that were obtained, will be discussed below. Regarding the field of view (FoV) determination experiment, it should be noted that the experiment has been carried out with three different objects, in particular: a round shaped metal stick, a round shaped plastic stick and a square shaped metal stick. In addition, it was experimented with a square shaped wooden stick, but the experiment could not be validated for the reason that its thickness was considerable and the FoV could not be determined with accuracy. Once obtained the results with the different point objects, it has been concluded that the material does not affect the FoV determination, but the form alters the result seeing that for a square shape better results have been obtained. This is because the echo of the signal sent by the sensor is easier to receive when it bounces on a square object than a round one. The results achieved are the following: a horizontal FoV of 60 degrees and vertical around about 30 degrees and a maximum length of 2.6 m as well. About the FoV width , the results have been similar to what was expected, while the FoV length has been better than expected in previous studies. With respect to the experiment of maximum and minimum length, it has been viewed that the Bosch sensor is able to detect objects at a distance of 4 meters, but is not very reliable since at these distances it is not able to detect the object at 100% of the time. Therefore, to achieve 100% effectiveness, the maximum distance detectable by the sensor is reduced to 3.6m. Note that in this experiment a greater maximum length has been obtained than in the FoV experiment because in this experiment the setup has been placed in front of a wall and not a point object, so that the wall is easier to detect due to its size. On the other hand it has also been demonstrated if these sensors have implemented the cross echo detection. The cross echo is the signal that comes from another sensor after it has bounced on an object. During the experiments, its operation has been confirmed, in such a way that with the cross echo functionality implemented, the whole sensor setup FoV improves. Finally, the calculation of the distance between object and sensor has been tested in such a way that the result obtained was as accurate as possible. This has been achieved without any problem reading the communication signal between sensor and control unit. Later, the script implementation for Arduino was carried out, which was able to calculate the distance to an object in real time without having to observe the communication Bus signal. Initially, the results obtained had an measurement failure about 3 cm with respect to the actual measurement approximately, even though this failure was not always obtained in the measurement. To solve this error in the measurement, it has been proposed to calculate the average of N results to get a mean value measured and thus improve the measure accuracy, reaching a maximum error of 1 cm. This mean calculation implementation is used to get a better sensitivity as well. To sum up, remark that the Bosch sensors are very powerful which is not reflected in the user-level functionality. It is because a parking assistance system does not require this accuracy and these wide ranges of vision. On the other hand, a system for autonomous driving need it, and according to the results obtained, Bosch sensors are good to do it. Emphasize that all the information or results collected Version: September 17, 2018 References Arduino IDE. Arduino IDE. URL: https://www.arduino.cc/en/Main/Software. Arduino UNO Rev3. Arduino UNO Rev3 Tech Specs. URL: https://store.arduino.cc/arduinouno-rev3. Arithmetic Mean Formula. URL: https://en.wikipedia.org/wiki/Arithmetic_mean. Bosch Operating Intructions. Bosch operating intructions. 2009. Bosch Ultrasonic Sensor. URL: https://www.bosch-mobility-solutions.com/en/products-andservices/passengercarsandlightcommercialvehicles/driverassistancesystems/ construction-zone-assist/ultrasonic-sensor/. Carullo, A. et al.: Ultrasonic sensor distance measurement (2001) Carullo, Alessio; Parvis, Marco: Ultrasonic sensor for distance measurement in automotive applications, in: IEEE Sensors journal, 1, p. 143, (2001) Data acquisition. Data acquisition. URL: https://www.ni.com/data-acquisition/d/. Frequency and Period Time relation. URL: https://en.wikipedia.org/wiki/Frequency. Fu, J.: Diss., Ultrasonic sensor model (2018) Fu, Junfeng: Auswahl und Implementation eines Ultraschall-Sensormodells für die Fahrzeuganwendung, (2018) Fu, J. et al.: Diss., Setup for ultrasonic models validation (2017) Fu, Junfeng; She, Jiahao; Shen, Liwen; Wanf, Qin,; Zuo, Yueyao: Aufbau eines geeigneten Test-Setups zur Validierung von Ultraschallsensor-Modellen, (2017) Jaffe, B.: Piezoelectric ceramics (2012) Jaffe, Bernard: Piezoelectric ceramics, Elsevier, (2012) Kishonti, L.: Real and simulated tests (2017) Kishonti, László: In autonomous driving, real-world testing is taking a backseat, (2017) Krzentz, S. V.: Voltage divider (1998) Krzentz, Steven V: Resistance voltage divider circuit, Google Patents, (1998) Nakamura, K.: Ultrasonic applications (2012) Nakamura, Kentaro: Ultrasonic Transducers: Materials and Design for Sensors, Actuators and medical applications, Elsevier, (2012) Navarro: Ultrasonic distance measurement (2004) Navarro: Sensores de Ultrasonido usados en Robótica Móvil para la Medición de Distancias, in: Scientia et Technica, 273, (2004) Olympus, N.: Ultrasonic Transducers (2006) Olympus, NDT: Ultrasonic transducers technical notes, in: Technical brochure: Olympus NDT, Waltham, MA, pp. 40–50, (2006) Real and Simulation car tests. URL: https://mashable.com/2017/07/14/autonomousdrivingreal-cars-simulations/?europe=true#FSNdrlOXsgqS. Rubio, C.: Ultrasonic transducer (2010) Rubio, Carlos: Fabricación de Transductores Ultrasónicos para Equipos automatizados de inspección de líneas de Tuberías, (2010) Rubio, C.: Diss., Ultrasonic transducer for automated sytems (2018) Rubio, Carlos: Transductores Ultrasónicos para Equipos automatizados, (2018) Winner, H.: Handbook of Driver Assistance Systems (2014) Winner, Hermann: Handbook of Driver Assistance Systems, ISBN: 978-3-319-09840-1, (2014) Yang, C. E.: Piezo electric effect (2016) Yang, Carmen Emily: What is the piezo electric effect?, (2016) Version: September 17, 2018