Evaluation on safety performance of a millimetre wave radar-based autonomous vehicle

CHEN Ding, NI Jin-ping, BAI Lang, CHEN Da-chuan

(1. Shaanxi Province Key Laboratory of Photoelectric Measurement and Instrument Technology,Xi’an Technological University, Xi’an 710021, China；2. School of Optoelectronic Engineering, Xi’an Technological University, Xi’an 710021, China；3. Faculty of Electronic Engineering, Aviation University of Air Force, Changchun 130022, China)

Abstract： This paper presents a method using range deception jamming to evaluate the safety performance of the autonomous vehicle with millimetre wave (MMW) radar. The working principle of this method is described. Combined with a waveform edition software, an experimental platform is developed to generate a deceptive signal that contains false distance information. According to related theories and its principle, the configuration parameters of the experimental setup are calculated and configured. The MMW radar of evaluated vehicle should identify an objective when it receives the deceptive signal from the experimental setup. Even if no obstacle, the evaluated vehicle can immediately brake in order that its braking distance is measured. The experimental results show that the proposed method can meet the requirements of the safety performance evaluation for the autonomous vehicle with MMW radar, and it also overcomes some deficiencies of previous methods.

Key words： vehicle safety evaluation； braking distance measurement； semi-physical simulation； range deception jamming； spurious echo generation

0 Introduction

With the development of artificial intelligence and automotive industry, automatic driving technology has become a research hot topic in many countries. However, road safety issue is a critical factor to affect the popularization of this technology. That is a prime reason why scientists and engineers are working on its applications, trying to develop more advanced vehicle emergency braking systems to protect driver and passengers from car collisions. For greater safety, multi-sensor information fusion theory is widely used in advanced vehicle[1]; therefore, this vehicle must employ various sensors based on the different detection principles, such as ultrasonic[2], infrared[3-4], laser[5-6]and millimetre wave (MMW) radar[7-8]. In comparison with other sensors, MMW radar has some advantages, that is, long detection range, due to the physical properties of MMW[9]. Most importantly, MMW radar cannot be significantly affected by weather whether in the day or at night, except for extremely bad weather (i.e. rainstorm, snowstorm, hail, etc.)[10]. For the above reasons, MMW radar can provide a timely and efficient warning for high-speed vehicle. Considering transport security issues, it is thus necessary to evaluate the performance of the vehicle emergency braking system with MMW radar effectively and reliably.

Until now, many evaluation methods have been proposed in this field. We can classify broadly these methods into three categories, including complete vehicle test[11-12], independent radar test[13]and fully numerical emulation[14-15]. Usually, the complete vehicle test is the most common evaluation method. Researchers set an obstacle on the route of a vehicle under test in advance, and then the vehicle is driven toward this obstacle along the specified route at a certain speed. Certainly, the results are most reliable and closest to reality in the complete vehicle test. Unfortunately, the functions of a new vehicle are not perfect in an early phase. Chances are that the evaluated vehicle is destroyed and severely damaged in the test. In independent radar test, the MMW radar is detached from the vehicle to evaluate its detection performance. It is unacceptable because its detection performance cannot represent the safety performance of the evaluated vehicle completely. Besides, fully numerical emulation is carried out in computer. Obviously, the reliability of results in the latter two methods is very poor. In addition, a method based on semi-physical simulation can effectively evaluate the evaluation of a photoelectric detection system used in the measurement of small arms (e.g. pistol, rifle, machine gun, etc.)[16]. Particularly, it overcomes the deficiencies of traditional methods (e.g., live ammunition test and fully numerical emulation). In addition, according to a news report in China, several hackers can adopt an MMW radio transmitter to stop a moving autonomous vehicle. Unfortunately, this news report did not describe any technical details, even system structure. According to the evaluation requirement of the autonomous vehicle with MMW radar, the proposed method combines the semi-physical simulation with those hackers’ technique in order to avoid the drawbacks of previous methods.

In this paper, it is presented that the working principle of the method and analysis of its feasibility. According to the theory of range deception jamming, the waveform of an echo signal with user-defined range information should be created. The SystemVue software is used to edit the waveform file of the radar signal, and then this file is downloaded into a semi-physical simulation platform to generate the radar echo. The purpose is to adopt the above platform to trigger the emergency braking system of the measured vehicle through its MMW radar. Provided the user-defined range information is less that the minimum safe distance of evaluated vehicle, the vehicle will stop immediately when its MMW radar receives a radar echo. In other words, the MMW radar is deceived to falsely identify a non-existent obstacle whose range relative to the evaluated vehicle is less than minimum safe distance. Therefore, the braking distance of the vehicle with MMW radar can be accurately measured to assess its emergency braking performance. The proposed evaluation method can meet the requirements of the safety performance evaluation for the autonomous vehicle with MMW radar, and it also can overcome the deficiency of previous methods.

1 Method

1.1 Working principle

Generally, MMW radar adopts frequency modulation continuous wave (FMCW) ranging method[17]. Compared to the ranging of pulse radar, the main advantages of this method are its ability to have the higher accuracy in a short range, the lower peak power of the transmitted signal, relatively simple circuit and small size[18]. Therefore, the FMCW method is an optimum choice for MMW radar. In time-frequency distribution, FMCW signal usually has several modulation waveforms, such as triangular, sawtooth and sinusoidal. When a transmitted signal with the given modulation waveform meets an obstacle, it is reflected immediately; in particular, its echo also has the identical modulation waveform, but this echo is only delayed for a time Δτrelative to the transmitted signal. In MMW radar, the frequency difference Δfbetween the transmitted signal and the echo is obtained through mixing. According to the principle of FMCW ranging[17], MMW radar can accurately measure the range from an obstacle to the vehicle if the accurate frequency difference Δfis obtained.

Currently, FMCW ranging with triangular waveform and that with sawtooth waveform are used most extensively in all kinds of MMW radar. Triangular waveform is regarded as a combination of the two sawtooth waveforms with identical shape but different orientations. Considering representativeness and supposing that an MMW radar adopts the FMCW ranging with triangular waveform and there is an obstacle in the driving direction of the vehicle under test, the time-frequency distribution curve of its transmitted signal and its echo is shown in Fig.1. In the upper part of Fig.1,fois the centre frequency of the transmitted signal frequencyft.TMis the frequency modulation period, and it is usually tens of milliseconds.Bis the frequency modulation bandwidth. In the interval of[fo+B/2,fo-B/2],ftvaries directly with the former halfFMperiodTM/2 and inversely with the latter halfFMperiodTM/2.fris the frequency of the received signal; obviously, this time-frequency distribution is identical with that of the transmitted signal. In the lower part of Fig.1,fbis the absolute difference between the transmitted signal and the received signal. This absolute difference is also called beat frequency.

Fig.1 Time-frequency distribution curve of transmitted signal and received signal

According to Ref.[19], the measured rangeRis obtained by

R=fbTMBc/4

(1)

with

fb=2BΔτ/TM

(2)

and

Δτ=2R/c,

(3)

wherecrepresents the light velocity, and Δτrepresents the delay time between the transmitted signal and received signal.

To ensure a single ranging result,TMmust meet the following condition， namely

TM?Δτ.

(4)

Provided that the distance from an obstacle to the vehicle is very far so that delay time is larger than theFMperiod, that is, the actual delay time Δτ′ can be expressed by

Δτ′=KTM+ΔτF, (K=1,2,…,N)

(5)

with

ΔτF=Δτ?TM,

(6)

where ΔτFis the delay time corresponding to false range.

Because the time-frequency distributionfr(t) is the periodic function withTM, calculated delay time is notKTM+ΔτFbut ΔτF.

Through the above analysis, the several different ranges may be measured to get an identical result in theory. In pulse radar, this problem is referred to as “range ambiguity”[17]. Generally,TMis tens of milliseconds. In theory, its maximum unambiguous range is even above a thousand kilometres provided that neglect of propagation loss. In fact, the propagation distance of MMW radar is usually limited due to its propagation characteristics. According to the current report, the detection range of MMW radar is generally above 300 m or less. Thus, there is not the problem of range ambiguity in the application of MMW radar.

However, the problem of range ambiguity can be purposely taken advantage to realize range deception jamming. This method is as follows: a deception jammer is placed outside the maximum detection rangeRDof the MMW radar. This makes sure that the deception jammer cannot be detected by the MMW radar. However, MMW radar can be successfully deceived by the spurious echo from the jammer. Most importantly, this deception jammer can generate the spurious echo whose delay timeτmust meet Eq.(7), which is given by

KTM≤τ≤KTM+τA, (K=1,2,…,N)

(7)

with

τA=2RA/c,

(8)

whereRAis the minimum alarm distance, andτAis the delay time corresponding toRA.

When the MMW radar receives the spurious echo, it can be deceived to deem its range relative to this obstacle less than the minimum safety distance. Next, a warning signal is sent to the emergency braking system in the vehicle.

Note that the distance from the deception jammer to the MMW radar of the evaluated vehicle must be beyond the maximum detection rangeRD; otherwise, the evaluation method using deception jamming is unavailable. It is the same as the conventional complete vehicle test because the MMW radar of the evaluated vehicle identifies the deception jammer as an obstacle. As is known to all, the propagation distance of the jamming signal is far longer than that of the radar signal due to no reflection. Therefore, in order that the radar can be successfully deceived by the range jamming signal, the distance between the two both should be less than the maximum jamming range of this jammer. According to the radar equation[19]and jamming equation[20], the maximum detection rangeRT_Maxof the radar and the maximum jamming rangeRJ_Maxof the jammer are expressed by Eqs.(9) and (10), respectively, namely

(9)

wherePtis the transmitted power of the radar,Gtis the antenna gain of the radar,λis the wavelength of the radar signal,Pris the power of the echo received by the radar,Lis the total of the related losses such as system loss, propagation loss and ground plane loss.

(10)

wherePJis the transmitted power of the jammer,GJis the antenna gain of the jammer,Gt(θ) is the antenna gain of the jammed radar at the jamming directionθ,λis the wavelength of the jamming signal (or the radar signal),γJis the polarization mismatch loss coefficient between the jam antenna and the radar antenna,PrJis the power of the jamming signal received by the radar,LJis the total of the related losses such as system loss, propagation loss and ground plane loss.

Assuming that the antenna of the jammer aims at the jammed radar,GJ=Gt(θ)=Gt,Pt=PJ,Pr=PrJandL=LJ, we can obtain the relationship betweenRJ_MaxandRT_Maxas

(11)

For better jamming effects, the parameters of the jammer are purposefully set to be superior to those of the jammed radar. Obviously, the actual maximum jamming range is larger than the above theoretical value.

Therefore, the measured rangeRTJfrom the jammer to the measurement position should meet the following condition as

RMax

(12)

Provided that the distance between the radar and the jammer meets the above condition, the MMW radar cannot detect this jammer at all, but this jammer can deceive the MMW radar successfully.

For the sawtooth waveform, its time-frequency distribution curve can be considered the former half frequency modulation period of the case of triangular modulation waveform. Thus, its ranging principle is identical with that of triangular waveform. In contrast to the mode with sawtooth waveform, the mode with triangular modulation waveform can also measure the velocity of the evaluated vehicle relative to an obstacle. Except for special requirements, velocity measurement is just an additional function because every vehicle itself has a speedometer. In addition, FMCW ranging with sinusoidal waveform is not usually used in vehicle but altimeter. Despite higher accuracy, it is difficult to extract velocity and range information relative to other methods. More discussions about FMCW ranging and other related work can be found in Ref.[17] therein.

1.2 Feasibility analysis

In the work, it is a key question whether the evaluated vehicle can immediately perform emergency braking after its MMW radar receives the spurious echo. In fact, there is no obstacle in the detection range of the MMW radar. Even though the method is feasible, it is of no significance if its low success probability. Generally, if the level of a received signal is higher than a given fixed threshold (or dynamic threshold), it should be considered the signal reflected from a target; or else, it should be classified as noise or clutter. In radar signal detection, an adaptive signal detection method, named constant false alarm rate (CFAR), is widely employed to obtain its high detection rate. The CFAR method is a binary Bayesian classification based on Neyman-Pearson detection criterion[21].

Assuming thatxis normalization observed signal (i.e. received signal),H0andH1represents no target and existence target, respectively, based on the hypothesis that the state ofH0is true, its model is given by

H0∶xk=nk, (k=1,2,…,N),

(13)

Likewise, the model ofH1is also given by

H1∶xk=a+nk, (k=1,2,…,N),

(14)

whereais deterministic signal (i.e. signal with target information) and it is greater than 0.

Through multiple observations, we can obtain anN-dimension observed signal vectorx=(x1x2…

xN) whose elements are independent identical distribution Gaussian discrete variables. Based on the hypothesis that the state ofH0and that ofH1are true, their probability density functions (PDFs) are respectively given by

(15)

and

(16)

According to the theory of CFAR, we can use Lagrange multiplierμ(μ≥0) to formulate the objective functionJas

J=P(H0|H1)+μ[P(H0|H1)-a]=

P(H0|H1)+μ[1-P(H0|H1)-a]=

(17)

whereP(H0|H1) represents the probability whose judgement result is non-target despite the existence target,P(H1|H0) represents the probability whose judgement result is existence target despite non-target,P(H1|H1) andP(H0|H0) represent correct judgment probability,R0andR1are the 0 domain and the 1 domain, respectively.

To meet the constraint condition (i.e.P(H1|H0) is close toaas far as possible),μmust be reasonably chosen to minimize the objective functionJ. The geometric significance of the above process is intuitively shown in Fig.2. Usually,P(H1|H0) is also called false alarm probability, and it is set to a quite small value ofα(e.g.α=10-6). Through the CFAR processing, the thresholdμis constantly adjusted according to Neyman-Pearson detection criterion[21]. Consequently, the hatched area corresponding toP(H1|H0) can be constant and quite small, but that ofP(H1|H1) can be large as far as possible.

Fig.2 PDF of observed signal, division of decision domain and its decision probability

The above signal detection is performed under normal condition (or non-antagonistic environment). Relatively speaking, the evaluation is performed under antagonistic environment because many so-called echoes are from the jammer. In theory, a larger value ofaleads to the larger distance between the peaks of these two PDFs in Fig.2, Obviously, it is a big probability event that the vehicle adopts the emergency braking after its radar receives the jamming signal. For better jamming effects, the transmitted power of the jammer should be as higher as possible, but it cannot lead to the saturation of receiver in radar. In the experiment, the success probability is above 90%.

1.3 System components

In military application, distance deception jamming is widely regarded as an effective method for radar countermeasure[20]. When jammed by this method, the jammed radar likely obtains the false information of enemy objective; specially, measured distance is not the real distance between them. Facing this radar countermeasure method, commercial radar is almost defenseless. If they have the ability of counter-counter measure, their hardware and software system must be improved completely. Usually, the price of commercial radar (e.g. anti-collision MMW radar) is from tens of dollars to thousands of dollars. However, the cost of a radar with excellent anti-radar countermeasure is up to millions of dollars. Evidently, vehicle manufacturers are difficult to accept this MMW radar because its cost is far higher than that of the vehicle itself. Therefore, commercial MMW radar does not have the anti-electronic countermeasure ability.

In the work, the proposed semi-physical simulation platform can be considered a jammer. Referencing the scheme in Ref.[16], the improved semi-physical simulation platform is proposed, as shown in Fig.3.

Fig.3 Block diagram of proposed semi-physical simulation platform

The platform consists of several main components: antenna, low-noise amplifier (LNA), local oscillator (LO), up conversion (UC) module, arbitrary waveform generator (AWG, Keysight M8195A), PC, etc. Its working process is as follows. The key parameters (i.e.fo,BandTM) of the MMW radar of the evaluated vehicle are given by its manufacturer, as listed in Table 1. First, we can operate the SystemVue software to compile and create the waveform file of the jamming signal in the PC. Next, the waveform file is downloaded into the AWG via LAN, and the AWG can generate an intermediate frequency (IF) signal relative with the jamming signal. Most importantly, this AWG must be configured as the working mode of external trigger in advance; that is, this signal source can output the IF signal after its trigger port receives a trigger signal. According to experimental requirements, the parameter of trigger delay time is configured under this working mode. Thus, this can simulate the spurious echo which is delayed a certain time relative to the transmitted signal. Besides, it must be synchronization between the clock of this AWG and that of the jammed radar to obtain high-precision delay time. Through wireless transmission, the clock signal of the jammed radar is input into the external clock port in the AWG. Furthermore, the IF signal is converted to a radio frequency (RF) signal using the UC module combined with the LO. Finally, the RF signal is amplified by an LNA, and it is emitted at the direction of jammed radar via the antenna.

Table 1 Main parameters of evaluated MMW radar

1.4 Waveform edition

As mentioned above, a software named SystemVue is used to compile and create the waveform file of the corresponding jamming signal in the PC. As shown in Fig.4, the jamming signal model is composed of several special modules, including CW signal generator, signal downloader and other two data sink models (i.e. “123” module). Most importantly, the function of the FMCW signal generator is to create a base band (BB) signal of FMCW with the user-defined features. According to the signal features of the measured MMW radar, we can select a reasonable waveform type and configure related parameter items in the module, as indicated in Table 1. Note that the “LowerFreq” item is 0 kHz due to BB signal. In this module, there are two different output ports. Here, the lower one can transmit the BB signal into the next module. Next, the BB signal is modulated to an IF signal with the carrier frequency of 33 GHz in the modulator module. Usually, IF is about half of carrier frequency. In the study, a UC module with the frequency range of 44 GHz is adopted. Therefore, the IF is set to 33 GHz. Particularly, the carrier frequency is not the carrier frequency of the measured MMW radar signal, but the carrier frequency of the IF signal. Then Gaussian noise is added to the IF signal after it passes through the noise density module. This module can simulate a complex electromagnetic environment in order that the spurious echo is closer to reality. Thus, we adjust the level of the noise density to evaluate the detection performance of the jammed radar. Usually, the level of the noise density should be set to the minimum (i.e. -130 dBm).

Fig.4 Schematic diagram of jamming signal model in SystemVue software

Finally, the IF signal data are transmitted to the signal downloader. Most importantly, this module can compile the waveform data of the IF signal to download into the AWG via LAN. Compared to Matlab, there is an unparalleled advantage for the SystemVue software; that is, it cannot only perform the numerical simulation and analysis in PC, but also generate an actual signal in the input port of the AWG. Therefore, we can employ this software to conveniently generate the actual IF signal according to the testing requirements. Note that the SystemVue Software combined with the related instruments of Keysight can only perform the function.

Through waveform edition, the IF signal with 33 GHz can be accurately generated in the output port of the AWG. The IF signal with 33 GHz is changed to RF signal with 77 GHz via the UC module. Next, the RF signal with 77 GHz is amplified using LNA, and then it is transmitted to antenna. Finally, the antenna can eject the jammer signal with 77 GHz to assessed vehicle.

2 Experiments

2.1 Experimental setup

According to Ref.[22], the test should be performed on the smooth, firm, clean and dry concrete or asphalt pavement; especially, the attachment coefficient between the pavement and the tire of the vehicle is greater than or equal to 0.7. In this work, the size and total no-load weight of the vehicle are 2 695 mm (L)×1 663 mm (W)×1 555 mm (H) and 920 kg, respectively. The width of the pavement should be 2.5 m if the width of the vehicle is less than 2.55 m. Moreover, the velocity of the measured vehicle should be 50 km/s if its weight is less than 3 500 kg.

The experimental setup is shown in Fig.5. Supposing that the vehicle travels along the pavement at a velocity of 50 km/s, the semi-physical simulation platform (i.e. jammer) can receive a trigger signal as soon as this vehicle passes through the detection area of a pair of infrared switches (also known as trigger point). Immediately, the jammer emits a beam of jamming signal toward the ahead of the vehicle. Despite there is not an obstacle on the road ahead, the vehicle is deceived to adopt the emergency braking after its MMW radar receives the jamming signal. Finally, the vehicle stops at a position completely; here, the position is called termination point. We can measure the range from the trigger point to the termination point to obtain the braking distance. Besides, the jammer is mounted on a door-shaped trestle whose height (2 000 m) is greater than that (1 555 mm) of the vehicle. This can prevent the signal generator device from being crushed by the vehicle whose emergency braking system is invalid, and this uncontrollable vehicle can pass through the underside of this door-shaped trestle properly.

Fig.5 Schematic (top view) of experimental setup

Prior to experiments, we should calculate some configuration parameters scientifically and reasonably. For the measured MMW radar, its maximum detection rangeRT_Maxand minimum alarm rangeRAare 150 m and 15 m, respectively. Additionally, the polarization mismatch loss coefficientγJbetween the jam antenna and the radar antenna is also measured in advance, andγJis approximately equal to 0.5. Substituting the above known parameters (i.e.RT_Max=150 m andγJ=0.5) into Eq.(10), we can obtain the maximum jamming range of the jammer, i.e.RJ_Max≈15 910 m, theoretically. According to Eq.(11), the distance should be in scope from 150 m to 15 910 m. However, we should take into consideration a few factors, such as the length limitation of the test pavement, the difficulty of installing this experimental equipment and the controllability of the experiment. Consequently, this pair of infrared switches is placed in the position where the distanceRTJrelative to the jammer is 1 500 m. Even if the jamming signal is feasible, the measured vehicle does not immediately adopt the emergency braking after the infrared switches are triggered. This is because that there is transmission time of trigger signal and jamming signal. Main configuration parameters of the experiment setup are listed in Table 2.

Table 2 Main configuration parameters of experiment setup

However, the total of the two delay times is only dozens of microseconds. For braking distance measurement, the impact of the above two delay times can be negligible. However, this may affect the feasibility of the experiment if the extra delay is added to the delay time of jamming signal. As a result, the vehicle does not adopt emergency braking because its MMW radar cannot be successfully deceived. Therefore, the transmission delay should be compensated. Based on the above analyses, the trigger delay time configurationτAWGin the AWG is given by

τAWG=TM+τ-2RT/c,

(18)

with

0≤τ≤2RA/c,

(19)

whereτis the delay time corresponding to the false range.

2.2 Experimental results

In outdoor, the propagation characteristics of the synchronous signal may also lead to the phase deviation between the clock of the measured radar and that of the jammer. Generally, the phase deviation obeys a uniform distribution with 0 mean, and it is negligible. Based on the above theoretical calculation, the static tests of jamming effect should be also carefully performed the prior to dynamic evaluation experiment. This aim is to obtain the reasonable configuration of the trigger delay time and transmitted power. Specially, these assessments cannot be carried out in succession too many times. This probably leads to super-heat brake block so that emergency braking is failing. According to experience, a set of assessments cannot be thus performed over ten times. Based on the safety considerations, in several evaluations, desired result is a maximum value but not a mean value (or variance, confidence interval, etc.).

After everything is ready, the maximum braking distance of a vehicle was measured under four different basic pavement conditions according to national technical standards in China[22], and corresponding evaluation results are listed in Table 3. Note that all results in Table 4 are up to 2 decimal places. According to the related criteria as illustrated in Table 3, we can get the objective and scientific conclusion; that is, the safety performance of the evaluated vehicle is in full with the national technical requirements of China[22].

Table 3 Results of safety performance for a vehicle

Table 4 Chinese national criteria of emergency braking system performance

Maximum braking distance under no-load (m)Maximum braking distance under full-load (m)Pavement condition≤21≤22Ordinary≤42≤44Wet≤63≤66Snowy≤123≤133Icing

Based on the same pavement conditions, the performance evaluation of the same vehicle was also performed using the complete vehicle test. Some obstacles (i.e. wooden cases, dummies and traffic barricades) were placed on the forward route of the measured vehicle. Fortunately, no obstacle was smashed by the vehicle during these experiments. Once, an uncontrollable vehicle was seriously damaged after it hit a concrete obstacle. The evaluation results based on complete vehicle test are given in Table 5. The evaluation results are also in accordance with the requirements of national technical standard[22]in China.

Table 5 Evaluation results of emergency braking system performance in complete vehicle test

Maximum braking distance under no-load (m)Maximum braking distance under full-load (m)Pavement condition1518Ordinary3034Wet4955Snowy96103Icing

3 Conclusion

In this research, a novel method is proposed to evaluate the safety performance of autonomous vehicle with MMW radar. Above all, the rationality of its working principle using distance deception jamming is analyzed scientifically. Next, the signal generator device can generate the spurious echo that contains false distance information. In the experimental setup, some related parameters of the MMW radar signal, such as delay time, jammer position and transmitted power, are reasonably configured according to its working principle. Then the vehicle can immediately adopt the emergency braking as soon as its radar receives the spurious echo. Consequently, the braking distances of the vehicle can be measured to evaluate its performance of the emergency braking system under different pavement conditions. Therefore, this method can avoid defections of the previous methods. Specially, it reduces the chance that the vehicle is destroyed in the evaluation experiment. Besides, the major components of the proposed equipment are all general-purpose instruments (i.e. AWG). These instruments are again used in the experiments for other radio equipment after this evaluation experiment. Obviously, development costs of a new vehicle can be greatly reduced in an early phase.

In future research, my team will focus on expanding the function of the experiment platform to evaluate the autonomous vehicle based on other sensors. For higher safety performance, it is our suggestion that vehicle manufacturers carry out other assessments under more complex pavement conditions. In addition, the method can also find the security holes of an intelligent vehicle effectively in order that the manufacturer of this vehicle can timely make up for them. Furthermore, the method can also prevent illegal intelligent vehicles from entering a protected area. Certainly, we also think it is illegal that a moving autonomous vehicle is arbitrarily jammed using the proposed method, and proposed legislation should prohibit this without permission.