Roe et al. (1992) [13] used FMCW radar (10.525GHz) to classify vehicle types, including bicycle, car, light goods, medium goods, heavy goods, and buses based on
17
Table 2.1: Literature review for vehicle classification based on radar-based systems
Authors Radar Type Features Installation
Roe and Hobson FMCW length & height over-head
Sang et al. FMCW BW & received power side-fire
M. Cherniakov et al. FSR received power side-fire
Ildar Urazghildiiev down-looking spread height profiles over-head
Jianxin et al. FMCW doppler signature over-head
length and height of cars, and accuracies of 75% in single lane are achieved. Park et al. (2003) [14] developed a K-band (around 24GHz) side-fire FMCW radar for a single lane application, and vehicles are classified as large, medium or small size based on received power and spectrum pattern. Cherniakov (2005) [6] use forward scattering radars (FSR) for classifying ground vehicles, but the transitive and receiving radar antennas are located at different sides of road. Principle component analysis is utilized to reduce dimensionality in frequency-domain feature vector, and the K-nearest is employed to form a classifier. The experimental cars are distinguished into three vehicle categories, including small, medium, and large. The experimental results show that recognition accuracy is only 71% when using the main lobe, while it keeps at 86% without the main lobe. Urazghildiiev et al. (2007) [5] classified vehicles based on vehicle height and length in a single lane, and this exhibits that the signals have the rough geometric pattern similar to the side view of vehicle. The key point of this study is to develop a real-time vehicle classifier for “road-side” equipments in “multi-lane environments”. (1) The road-side equipment is perpendicular to the traffic direction, which makes Doppler effect unapparent to be utilized. (2) The road-side equipment receives the signals of all lanes which requires more sophisticated model and algorithm to describe and analyze our data.
After reviewing the developments of vehicle classification, it is known that, for
real-world multi-lane applications, there are few significant results and field tests in vehicle classification based on road-side radar detectors. Moreover, using road-side radar detectors for classifying small and large vehicles in multi-lane environments simultaneously will lower the cost of acquiring the traffic information, consequently, this study develops mathematical modelling as a working tool based on FMCW radar data to analyze the vehicle types in multi-lane environments.
Chapter 3
Road-Side Radar System Design
Real-time systems comprises all devices with different constraints and limitations. In general, we can classify these constraints into two categories, that is,“Hard deadlines”
and “Soft deadlines”. Hard deadlines are constraints that absolutely must be met. A missed deadline constitutes an erroneous computation and a system failure. In these systems, late data are bad data. On the contrary, if late data are still good data, such a system is referred to a “soft” real-time system. In this study, when the term real-time is used alone, we specifically referring to hard real-time systems [15].
In addition, this study only focuses on the details of software design, thus the hardware design is out of our research scope. However, our proposed softwares are working on the certain requirements of hardware, so these requirements of hardware design will also be briefly introduced in the following section.
19
3.1 Hardware Architecture and Design Specifica-tions
An X-band CMOS-based frequency-modulation continuous-wave (FMCW) radar sys-tem is proposed by Tzuang et al. [16] and adopts two antennas to transmit and receive signal separately. The radio frequency (RF) transceiver is formed by standard 0.18 µm one-poly six-metal (1P6M) complementary metal-oxide semiconductor (CMOS) technology and is embedded in a chip area of 1.68 mm × 1.6 mm. The several sig-nificant properties of the FMCW radar are described as follows. First, the linearity of VCO is 3 % while the output frequency is ranged from 10.3 GHz to 10.8 GHz.
Second, the on-chip isolation between the transmitting and receiving paths is 55.0 dB at 10.5 GHz, additionally; two planar leaky-mode antenna arrays with a gain of 18 dB are designed. Experiments indicate the isolation between two antenna arrays with a spacing of 5.0 mm is higher than 42.0 dB at 10.5 GHz.
Figure 3.1 illustrates the block diagram of the X-band FMCW radar system, which comprise dual planar antenna arrays at the transmitter output and the receiver input; a 0.18 µm 1P6M CMOS transceiver is responsible for the RF signal processing as shown inside the dashed lines, and a baseband digital signal processing unit is designated for instantaneous and simultaneous assessment of range measurements.
Notably, using two antennas rather than only one eliminates the circulator, as is expensive and does not provide sufficient isolation between the transmitter and the receiver.
In the range measurement, the FMCW radar transmits the signals, and receives the reflecting waves from an intruding vehicle. The amplitude of the input triangular
21
Figure 3.1: Block diagram of the X-band FMCW radar system which is composed of two antenna arrays, a single-chip CMOS transceiver (enclosed by the dashed line) and an external digital signal processing unit. Power amplifier is added to increase the output power level [16]
wave directly controls the output frequency of the transceiver. The bandwidth of the modulated signals at the transmitting path, denoted by BW, is equivalent to the frequency difference between the maximum and minimum output frequencies of the voltage controlled oscillator (VCO). Moreover, the modulated-bandwidth over the half-period of the triangular-wave, denoted by Sb, can be calculated by
Sb = BW
tm/2 (3.1.1)
where tm denotes the period of the triangular wave, and the parameters are set when performing the signal processes. The pulse-repeated-frequency (PRF) is 500.0 Hz; the swept time (Ts) is set to 2.0ms and the period of triangle-wave tm is 0.5 ms.
The beat frequency, denoted by fb, is the frequency difference between two input
signals applied to the mixer. If the round trip time for signal propagation between the sensor and the target is less than that of a half-period of a triangular wave, then Sb can be maintained at a constant value. Suppose the maximum distance R between the radar and the target is set 60 meter, and the round trip time τ is described as
τ = 2R
c , (3.1.2)
where c represents the speed of light in air. Consequently, the maximum round trip time τ is 0.4µm. The beat frequency fb shown in (3.1.3) also can be obtained if any reasonable round trip time is given.
fb = 2 × BW
tm τ. (3.1.3)
On the contrary, the quantity of R also can be calculated by (3.1.4) if fb is given.
R = c × tm× fb
4 × BW . (3.1.4)
The ideal range resolution R0 of the FMCW radar can be estimated as
R0 = c
2 × BW, (3.1.5)
thus, range resolution R0 is 1.0 meter. Equations (3.1.1) to (3.1.5) illustrate the principles of FMCW radar for range measurements. These operations are based on the assumption that the frequency modulation is linear. In other words, the slope of modulation-bandwidth, Sb, must be maintained at a constant value during the half-period of the input triangular wave. If the output frequency of VCO is not lin-early proportional to the amplitude of input triangular wave, then the linear transfer function between fb and τ shown in (3.1.3), is not valid. Consequently, the range
23
resolution of the FMCW radar is degraded. Moreover, the signals at the transmitting and receiving paths are assumed to be independent of each other in the FMCW radar.
The coupling effects between the transmitting and receiving paths are not included in the above derivations, hence, such leakages could significantly reduce the performance of the FMCW sensor. The design issues, which include linearizing the generating fre-quency modulating signal and eliminating undesired couplings in the radar system, are significant challenges for the proposed single-chip CMOS transceiver.
Figure 3.2 illustrates the illuminative field of the designed antenna [16] from top and side views. The angle ψ shown in Figure 3.2(a) is called elevational plane beamwidth, and is asked to be large, because the larger angle ψ represents that more lanes can be covered. The angle θ shown in Figure 3.2(b) is called horizontal plane beamwidth, and is expected to be small. The design reasons for the horizontal plane beamwidth as θ can be described as follows. First, if the horizontal plane beamwidth displayed as a large angle θ0 shown in Figure 3.2(b), the information about vehicles in which lane can not be caught. Take vehicles A and B for example, vehicles A and B will form the similar frequency-domain information theoretically because both vehicles have the same relative distance to a detector. Therefore, the large horizontal plane beamwidth θ0 makes different vehicles’ reflected signals overlap in the same position such that vehicles in which lane can not be recognized. Second, consider-ing the farthest lane (relative to a road-side radar detector), if the horizontal plane beamwidth is θ0, the illuminative field in the farthest lane may contain more than one vehicle, such that road-side radar detectors can not separate vehicles one by one.
Take vehicles B and C for example; vehicles B and C will be misled as an only vehicle if the horizontal plane beamwidth is θ0, then the erroneous traffic count occurs in that
Table 3.1: Comparison of the current commercial products
RTMS Smart Sensor The designed
Model 105 product
Protocol RS-232 or RS-232 or RS-232
RS-485 RS-485
Power 4.5W 7.5W 5W
consumption
Temperature −37◦ ∼ 74◦C −40◦ ∼ 75◦C
Size 16 x 24 x 12 cm 32 x 23 x 7.6 cm 29.3 x 12.2 x 25.5 cm
Weight 2.2 kg < 2.27kg 3.5 kg
lane. In a word, the illuminative field of the antenna determines the position of the farthest lane.
Table 3.1 shows the comparison of the current commercial products with the proposed radar sensor in this research. There is still room for improvement to the same grade products. Besides, the quality of materials and the shell over the detector may affect antenna radiation patterns such that the farthest distance is unable to be detectable. Additionally, antenna radiation patterns of dual planar antenna arrays are shown in Fig. 3.3 and 3.4. The 3dB beam width of the antenna array was 20◦ in the E-plane, and the main beam with a gain of 18.0 dB is at 56◦ in the H-plane.
25
(a)
(b)
Figure 3.2: (a) A side view of road-side radar detector in multi-lane circumstances.
(b) A top view of road-side radar detector in multi-lane circumstances. There are a small angle θ and a large angle θ0.
Figure 3.3: E-Plane (Microstrip Leaky-EH1-Mode Antenna Array) [16]
27
Figure 3.4: H-Plane (Microstrip Leaky-EH1-Mode Antenna Array) [16]
Node1
Figure 3.5: The C code is deployed on a DSP device, while the java code is deployed on a PC. The devices are connected by RS232, and communicated by the designed interface.