Po-JuiChiuandHsin-MuTsai AnIntersectionCollisionAvoidanceSystemforScootersUtilizingNon-Line-Of-SightLinks

(1)

An Intersection Collision Avoidance System for Scooters Utilizing Non-Line-Of-Sight Links

Po-Jui Chiu and Hsin-Mu Tsai

Department of Computer Science and Information Engineering National Taiwan University, Taipei 10617, Taiwan

E-mail:{r00922088, hsinmu}@csie.ntu.edu.tw

Abstract—In Taiwan, accidents involving scooters have con- tributed to more than 70% of the total number of injuries and fatalities in traffic accidents and the numbers have been increasing for the past 8 years; it is therefore crucial to derive cost-effective safety solutions which can be used in scooters.

To this end, we propose a cost-effective collision avoidance system which utilizes non-line-of-sight (nLoS) communication links between vehicles on intersecting roads. IEEE 802.15.4 link measurements were carried out to investigate the path loss and the packet reception rates of those nLoS links in three different intersections. Analysis of the results suggests that nLoS links at intersections, while exhibiting high path loss, are still more than sufficient to support the required vehicle-to-vehicle (V2V) communications for the proposed system to avoid 99% of the possible collisions at intersections.

Index Terms—Link Measurements, Scooter Networks, Vehic- ular Ad Hoc Networks (VANET).

I. INTRODUCTION

In the past 10 years, the development of both passive and active safety systems in cars has significantly reduced the number of fatalities and injures in traffic accidents; systems such as airbags, anti-lock braking system, electronic stability control, blind spot information system, and pedestrian de- tection system, etc., have become standard features in many new cars. However, the same sophisticated safety solutions could not be applied to scooters, which are widely used for personal transportation in many countries, such as Italy, Japan, China, and Taiwan. This is due to their low cost - an average scooter costs only about 2,000 U.S. dollars, approximately one tenth of the cost of an average mid-size passenger car.

In addition, scooters have almost no passive protection to the rider and thus the accident is often catastrophic when they are involved. Statistics show that this has become a significant problem: for example, in Taiwan more than 70% of the total number of injuries and fatalities due to traffic accidents are now contributed by those involving scooters [1].

To make sense for scooters in terms of cost, it is believed that vehicle-to-vehicle (V2V) communications is a key com- ponent in the solution, where vehicles, including cars and scooters, could periodically broadcast their current velocities, locations, headings, and sensor information from the on-board sensors to their neighboring vehicles. Without the need to add many costly sensors to a scooter, its rider could be more aware of the surrounding via V2V communications, and, in turn, lower the probability of an accident.

Fig. 1. Block diagram of the experimental setup

Fig. 2. Packet format

To this end, we propose a simple intersection collision avoidance system for scooters in this paper. At intersections, buildings at the corners often block the view of the drivers/riders to the side, preventing them from seeing the vehicles about to cross their paths. The drivers/riders thus enter the intersection blind to any approaching vehicles, leading to disastrous collisions. Moreover, these buildings could also block the Line-of-Sight (LoS) propagation paths, resulting in none-Line-of-Sight (nLoS) links between vehicles driving on two intersection roads and communicating with each other.

NLoS links are often either completely ignored or modeled in an overly simplified way as links with reduced range and, as a result, protocols often do not take full advantage of them.

In the proposed system, we utilize the nLoS links for direct Globecom 2012 - Ad Hoc and Sensor Networking Symposium

(2)

(a) Location 1 (b) Location 2 (c) Location 3 Fig. 3. Aerial photographs of the measurement locations (courtesy of Google Maps)

communications between vehicles on intersecting roads. On the other hand, we choose to use IEEE 802.15.4 [2] radios operating at 2.4 GHz for the proposed system because the technology is low-cost, more mature, and readily available compared to IEEE 802.11p [3] radios operating at 5.9 GHz.

The results are also applicable to other 2.4 GHz technologies, such as IEEE 802.11b/g/n (WiFi) technologies. We believe these 2.4 GHz technologies are more feasible for use in scooters due to their low cost. Moreover, the path loss at 2.4 GHz is generally much less than that of 5.9 GHz.

There have been several attempts to understand how nLoS links behave at intersections. Empirical data has been obtained in [4–7], but most of them are carried out at 5.9 GHz and the amount of measurement data is sometimes limited to a few transmitter-receiver location pairs. [8] derived several analytical propagation formulae to predict the nLoS path loss in the urban street grid microcellular scenarios. [9] evaluated a similar collision avoidance system utilizing the 200 MHz frequency band, but with simplified theoretical channel models.

This paper has two main contributions. First, to better understand nLoS link behaviors at 2.4 GHz, we carry out an extensive series of link measurements utilizing IEEE 802.15.4 radios at three different intersections in the urban area with different conditions. Second, we evaluate the feasibility of the proposed collision avoidance system by estimating the probability of avoiding possible collisions at intersections;

the analysis is based on realistic assumptions and utilizes our empirical measurement data. The rest of this paper is organized as follows. Section II describes the setup of the measurements, followed by section III, which presents the measurement results and their implications. We utilize the results to evaluate the feasibility of the proposed system in section IV. Finally, concluding remarks and future works are given in section V.

II. EXPERIMENTALSETUP

A. Experimental Hardware and Software

FireFly v2.3 [10] is used as our hardware platform in the experiments and has an on-board IEEE 802.15.4 radio, CC2420 [11]. In the experiments, both transmitting nodes

(TN) and receiving nodes (RN) use omnidirectional rubber duck antennas with a peak gain of 8 dBi. The antennas are mounted on tripods, perpendicular to the ground, and located 70 cm off the ground to emulate the scenario that they are installed on scooters. Two AA alkaline batteries are used to supply the power. The firmware of the node is based on Nano- RK [12], which is a fully preemptive reservation-based real- time operating system (RTOS) designed for wireless sensor networks.

Figure 1 shows the experimental setup which includes one TN and several RNs. The firmware of TN and RN perform the following tasks, respectively:

• TN: (1) Periodically create packets with increasing sequence numbers (increased by one after each packet creation); and (2) broadcast the packets to RNs.

• RN: (1) Collect all received packets broadcasted by TN; (2) send the information of received packets to the external microSD card logger (OpenLog) via a universal asynchronous receiver/transmitter (UART) connection;

and (3) OpenLog stores all received information on the microSD card. The data on the microSD card can later be imported on a PC for further analysis.

B. Communication Parameters

• Transmission power: 0 dBm (1mW) in all experiments.

• Packet sending rate: TN is configured to send 1000 packets in 110 seconds. The sending rate (about 9.1 Hz) is sufficient for most vehicle safety applications.

• Packet format: The packet format used in our experi- ments is shown in Figure 2. The total size of the physical layer service data unit (PSDU) is 21 bytes. There is only one data field in the payload, which records the sequence number of the packets from TN.

• Channel selection: In order to mitigate the effects of the interference from common WiFi devices in the 2.4 GHz frequency band, channel 26, which occupies 2.4785 GHz - 2.4815 GHz, is used in the experiments. The channel does not overlap with the bandwidth occupied by any 802.11b/g channel; the closest channel of 802.11b/g, channel 11, occupies 2.451 GHz - 2.473 GHz.

(3)

TABLE I

CHARACTERISTICS OF THETHREEMEASUREMENTLOCATIONS Location (Lat./Lon.) Widths of Roads Buildings Traffic

1 (25.01903,121.541281) 15m, 10m Many Light

2 (25.019091,121.546284) 60m, 40m Few Heavy

3 (25.024781,121.543313) 30m, 20m Many Heavy

C. Performance Metrics

• Received Signal Strength Indicator (RSSI): RSSI rep- resents the total amount of energy received within the operating frequency range, including the energy from the intended signal, the interference, and the noise. CC2420 has a built-in RSSI which has a resolution of 1 dB and a dynamic range of about 100 dB. The RSSI value is averaged over 128 µs. The typical sensitivity of the CC2420 radio is -95 dBm, which is approximately the lowest value which can be reported by RSSI.

• Packet Reception Rate (PRR): In this paper, PRR of a particular link is defined as

PRR ≡ ncrc ok

ntx

, (1)

where ncrc okis the number of correctly received packets by RN and ntxis the total number of packets transmitted by TN. Packets which are not correctly received can mainly be classified into two cases: either the packet is completely not detected or received by RN’s radio due to low Signal-to-Noise ratio (SNR), or the packet is received but in error. The sequence number field can be used to identify the former case (which is usually due to cor- rupted preamble sequence or start-of-frame delimiter) by observing non-consecutive sequence numbers in received packets, while the hardware generated CRC-16 field can be used to identify the latter case.

D. Measurement Locations

As the objective of the measurement is to understand how nLoS links behave at the corners of intersections, we chose to carry out the measurements at three typical 4-leg intersections in the Taipei city, which are shown in Figure 3. TN is placed on one side of a corner at the intersection and RNs are placed on the other to measure the quality of the link. The nodes are placed every 5 meters, ranging from 0 to 40 meters from the intersection (with a few exceptions where it is not possible to place the nodes for measurements). Measurements are carried out for each TN-RN location pair and only one TN is configured to transmit at a time. The nodes are static for the whole duration of measurements¹. The angles between the two sides of the measured corners are 90^◦, 90^◦, and 97^◦, respectively. Characteristics of the three measurement locations are shown in Table I.

−1000 −90 −80 −70 −60 −50

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

RSSI (dBm)

PRR

Location 1 Location 2 Location 3 Typical Sensitivity Minimum Sensitivity

Fig. 4. PRR versus RSSI profile

0.6 0.8 1 1.2 1.4 1.6 1.8 2

40 50 60 70 80 90 100

Distance(m) with log scale

Path Loss (dB)

Location 1 (L1) Fitted line for L1 Location 2 (L2) Fitted line for L2 Location 3 (L3) Fitted line for L3 Location Exponent n Constant C

1 4.22 15.73 2 4.54 16.43 3 4.56 10.89

Fig. 5. Path loss versus T-R distance and fitted lines of (3)

III. MEASUREMENTRESULTS

A. RSSI versus PRR profile

Figure 4 shows the RSSI versus PRR profiles of all three locations. Each data point represents the data set from a particular TN-RN location pair and is plotted with the coor- dinate(µRSSI,PRR), where µRSSI is the average RSSI of all correctly received packets. Although the data points are from different locations, they all follow a similar trend; the points form a curve whose PRR gradually drops from 1 to 0 as the RSSI decreases and approaches the receive sensitivity².

Comparing the differences between the profiles of the three locations, one can observe that the dropping curves of location 2 and location 3 are slightly to the right of and wider than that of location 1. Both are due to the fact that in location 2 and location 3 there are more moving vehicles and pedestrians acting as reflectors and diffractors in the propagation paths;

they create variations in RSSI, usually referred as fading.

1As the speed of the scooters in the urban area are usually less than 40 km/h, especially when they approach the intersection, the additional link quality variation caused by the scooter’s movement is not significant. As a result, we use the static setting for the measurements.

2According to the definition in [11], the receive sensitivity is defined as the RSSI value with which the transmission will have a bit error rate (BER) of 10−2.

(4)

Fading, in turn, would result in random packet drops/errors.

Hence, with the same average RSSI value, the curves of location 2 and location 3 have lower PRR, and more outlying data points (those that do not follow the curve).

B. Path Loss Exponent

The amount of energy loss during propagation is referred as path loss. Path loss, P L, when including the antenna gains, can be calculated as

PL = Ptx− Prx (2)

where Ptx is the transmission power and Prx is the received signal power. In our experiments, Prx can be approximated by RSSI, as the interference power and the noise power are negligible compared to the power of the received signal power.

The path loss is related to the transmitter-receiver (T-R) distance in general, and, in the simplified form, it is given by P LdB= 10n log₁₀(d) + C (3) where P LdBis the path loss in dB, d is the T-R distance, n is the path loss exponent that indicates the growth of attenuation during propagation, and C is a constant which combines the effects of other factors such as antenna gains.

Figure 5 plots the path loss in dB with respect to the T-R distance in log scale. Fitted lines of (3), also shown in the figure, are derived with the minimum square error method and the values of the parameters are also shown in the figure. One can see that while the actual numbers of path loss are slightly different among different locations, the path loss exponents are quite similar, all above 4.2. This is expected, as in our experiments all TN-RN location pairs do not have LoS propagation paths except when they are less than 10 meters to the intersection.

C. RSSI and PRR versus the Distance to the Intersection In Figure 6, the surfaces shows either average RSSI or PRR values when TN and RN are at different distances to the intersection. The X and Y coordinates represent the distances from the TN and from the RN to the intersection, respectively.

In Figure 6(a)(c)(e), one can observe that in all three locations RSSI gradually decreases to the value of the receive sensitivity as the distances increase. However, one can see that the surfaces are rugged; there are often random drops or rises among consecutive TN or RN locations, representing different amount of path loss due to the multipath effects caused by the particular shape of the buildings and the objects near the intersections. Note that this spatial multipath effect is different from the time variation described in III-A. In Figure 6(b)(d)(f), one can observe that PRR also gradually decreases from 1 to 0 in all three locations, but less regularly in location 2 and location 3; the lower PRR in certain TN-RN location pairs are due to fading caused by the moving vehicles and pedestrians, as well as the aforementioned spatial multipath effects.

IV. SYSTEMEVALUATION

In this section, we utilize the measurement results to investigate the performance of a simple intersection collision avoidance system. The main idea is that each vehicle will be equipped with a low-cost safety system consisting of an IEEE 802.15.4 radio with an omnidirectional antenna similar to the one used in our experiment and a localization system (typically a Global-Positioning-System (GPS) chip) to provide the location, the heading, and the velocity information of the vehicle. The system periodically broadcasts beacons with this information³. When receiving beacons from neighboring vehicles, vehicles on the two intersecting roads can determine if they are in danger of a collision (possibly at an intersection with stop signs, without traffic signals, or with little traffic) and if they are, warnings can be given to the drivers/riders to avoid the collision.

Further investigation of Figure 6(b)(d)(f) reveals that, PRR is more than 50% when TN’s and RN’s distances to the intersection, (X, Y ), satisfy X + Y ≤ {65, 40, 45} m for the three measurement locations, respectively. This means that if the vehicles utilize only nLoS links, in the worst case scenario the vehicles would start to have a fairly usable link when both vehicles are 20 meters or less from the intersection. With a typical velocity of 40 km/h in the urban area, it still gives the vehicles about 1.8 seconds to receive information from the other vehicle and to react, which seems sufficient to avoid the collision. In the following, we use a detailed model to evaluate whether the proposed system is feasible.

The following are the assumptions used in the analysis:

1) There are two vehicles, A and B, both bounding for the intersection with fixed velocities vAand vB4.

2) Only packets which are sent when the vehicles are less than 40 meters away from the intersection will be received by the vehicles on another leg of the intersection.

This assumption is validated by the measurement results and simplifies our analysis.

3) The link is symmetric, i.e., PRR is the same for the link from A to B and for the link from B to A.

The objective is to formulate the probability that a vehicle can avoid the collision by using the proposed system.

The typical safety braking distance for a vehicle with a velocity of v, S(v), is given by

S(v) = vTreaction+ v²

2µg , (4)

where µ is the coefficient of friction, g is the acceleration of gravity, and Treaction is the reaction time it takes for the driver to notice the warning and apply the brake; the first term represents the distance the vehicle travels while the driver/rider reacts, and the second term represents the minimum distance required for the vehicle to brake to a stop. We use 0.8 as

3It would be even better if the system is equipped with an electronic map so that it will only broadcast the beacons when it is close enough to an intersection. However, this is not absolutely necessary.

4This is reasonable as the driver/rider tends to keep an ideal velocity if they are under the impression that no other vehicle also bounds for the intersection.

(5)

0 5 10 20 15 25 30 35 40 0

5 10

15 20

25 30

35 40

−100

−80

−60

−40

−20 0

TN’s distance to the intersection (m) RN’s distance to the intersection (m)

RSSI (dBm)

(a) RSSI v.s. the distances to the intersection, location 1

0 5 10 15 20 25 30 35 40 0

5 10

15 20

25 30

35 40 0

0.5 1

PRR

(b) PRR v.s. the distances to the intersection, location 1

5 10 15 20 25 30 35 40 5

10 15

20 25

30 35

40

−100

−80

−60

−40

RSSI (dBm)

(c) RSSI v.s. the distances to the intersection, location 2

5 10 15 20 25 30 35 40 5

10 15

20 25

30 35

40 0

0.5 1

PRR

(d) PRR v.s. the distances to the intersection, location 2

10 15 20 25 30 35 40 45 10

15 20

25 30

35 40

45

−100

−80

−60

RSSI (dBm)

(e) RSSI v.s. the distances to the intersection, location 3

10 15 20 25 30 35 40 45 10

15 20

25 30

35 40

45 0

0.5 1

PRR

(f) PRR v.s. the distances to the intersection, location 3 Fig. 6. RSSI and PRR versus distances to the intersection

the value for µ [13], representing the general condition that the surface of the road is dry asphalt and the tires are in fine condition, and 0.75 seconds as the value for Treaction [14].

In the case of a possible collision, A and B would reach the intersection at the same time. Let DA(t) and DB(t) be the distances of A and B to the intersection at time t, respectively.

Therefore, DA(t) and DB(t) can be related by DB(t) = DA(t)vB

vA

, (5)

where vA and vB are the velocities of A and B, respectively.

Assume that t = 0 when A starts broadcasting periodic beacons. A’s and B’s distances to the intersection when A sends the i-th beacon are given by

DA(i− 1

f ) = DA(0) −vA(i − 1)

f (6)

and DB(i− 1

f ) = DB(0) −vB(i − 1)

f = DA(0)vB

vA

−vB(i − 1)

f ,

(7) respectively; f represents the beacon sending rate of A.

As we only measure PRR for certain TN-RN location pairs, we use linear interpolation of the PRR values of the 4 closest TN-RN location pairs to approximate Pe(dT N, dRN)⁵, which represents the empirical packet error rate when TN’s and RN’s distances to the intersection are dT N and dRN.

When the beacon size, Lb, is not the same as the packet size in the measurement (21 bytes), the packet error rate is

5In location 2 and location 3, we did not measure some TN-RN location pairs when one of them is very close to the intersection (0 m or 5 m); in this case we use the PRR value obtained at the next measurement location (5 meters further away from the intersection) for interpolation, which is usually lower, so that the final probability can be a lower bound.

(6)

0 2.5 5 7.5 10 12.5 15 17.5 20 0

5 10 15 20 25 30 35 40 45 50

Velocity (m/s)

Minimum Beacon Sending Rate (1/sec)

Location 1 Location 2 Location 3

Fig. 7. Minimum beacon sending rate to reach a 99% probability of avoiding the collision with 64-byte beacon

adjusted by

Pe,Lb(dT N, dRN) = Pe(dT N, dRN)^Lb²¹ , (8) where Pe,Lb(dT N, dRN) is the adjusted empirical Lb-byte packet error rate, under the assumption that the bit errors in a packet are independent.

To avoid the collision, B must correctly receive at least one beacon before its distance to the intersection is less than S(vB) so that it can stop before entering the intersection⁶.

The number of beacons which can be sent by A and received by B before its distance to the intersection is less than S(vB) is therefore given by

ni= (DB(0) − S(vB))f vB

. (9)

The probability of avoiding the collision due to beacon transmissions from A to B with given vA, vB, and f is given by

Psuccess,A→B

= 1 −

ni

Y

i=1

Pe,Lb

DA(0) −vA(i − 1)

f , DA(0)vB

vA

−vB(i − 1) f

(10) .

Finally, since B is also broadcasting beacons to A, it would also be sufficient to prevent the collision if A correctly receives at least one beacon before its distance to the intersection is less than S(vA). Therefore, the final probability of avoiding the collision is given by

Psuccess= 1 − (1 − Psuccess,A→B)(1 − Psuccess,B→A) . To reach a higher probability of avoiding the collision, it is necessary to increase the beacon sending rate. Figure 7 shows the minimum required beacon sending rate to reach the a 99%

probability at different velocities. One can observe that with velocities less than 12.5 m/s (45 km/h) and 64-byte beacons,

6To obtain a lower bound, this is a stricter condition; in reality it just needs to slow down sufficiently to allow A to pass the intersection first.

the required beacon sending rate is only at 5 beacons per second.

If we consider the scenario that there are multiple vehicles located near the intersection, as they are all broadcasting beacons, packets could collide with each other if their transmission durations overlap. Here we assume each beacon is broadcasted as soon as it is generated (using an “ALOHA”- style Medium Access Control (MAC)), without carrier sensing and random back-offs. We also assume that the beacons sent by different vehicles are not synchronized, so that the events of packet collisions are independent.

The average probability of a packet collision is Ppc =Lbf

R , (11)

where R is the data rate in bytes per second. In our analysis, we set R to be 31.25 kB/s (250 kb/s), which is the raw data rate of the 802.15.4 radio.

With N vehicles located near the intersection (those which can receive and transmit with each other), the probability that no packet collision occurs is given by

Pno pc(N ) = 1 − (1 − Ppc)^N = 1 − (1 −Lbf

R )^N . (12) The probability of avoiding the collision, after taking into consideration the possibility of packet collisions, is given by

Pˆsuccess= 1 − (1−Psuccess,A→BPno pc(N ))

(1 − Psuccess,B→APno pc(N )) . (13) Figure 8 shows the probability that the system fails to avoid the collision when both vehicles are traveling at 45 km/h with different number of vehicles located near the intersection and different beacon sending rate in the third location. As the vehicles tend to slow down and the driver/rider is more cautious when there are many other vehicles on the same road, where the collision avoidance system is not useful, we only evaluate scenarios with up to 20 other vehicles. One can observe that increasing the beacon sending rate would not obviously increase the overall failure probability when the beacon size is less than 128 bytes; in these cases the additional failures due to packet collisions are less significant. When the beacon size is as large as 256 bytes, increasing the beacon sending rate would significantly increase the failure probability and in cases where there are more than 5 other vehicles close to the intersection, the probability would never be less than 1%.

However, as we estimate that the beacon size is less than 64 bytes⁷, this is not the case. In summary, in all applicable cases, the required beacon sending rate is reasonably at 5 beacons per second.

7Latitude, longitude, velocity, and heading, gathered from the GPS, is essential for this application. With each field using the double-precision floating point representation, 4 × 4 = 16 bytes is required. The beacons also need to include vehicle identifier (license plate number, 6 bytes for Taiwanese plates), protocol headers, and other system information. We estimate the size of the beacon, including the overhead from the physical layer (PHY) and MAC (17 bytes), would be less than 60 bytes. This is much more conservative than the assumption in [9], where they state 6 bytes of payload is required for similar purposes.

(7)

5 10 15 20 10⁻¹⁵

10⁻¹⁰ 10⁻⁵ 10⁰

Fail Rate (Log Scale)

Beacon Sending Rate (1/second) N=1 N=5 N=10 N=15 N=20 0.01

(a) 64-byte beacon

5 10 15 20

10⁻¹⁵ 10⁻¹⁰ 10⁻⁵ 10⁰

Beacon Sending Rate (1/second) N=1

N=5 N=10 N=15 N=20 0.01

(b) 128-byte beacon

5 10 15 20

10⁻¹⁵ 10⁻¹⁰ 10⁻⁵ 10⁰

Beacon Sending Rate (1/second) N=1

N=5 N=10 N=15 N=20 0.01

(c) 256-byte beacon Fig. 8. Probability of failing to avoid the collision with different beacon size

V. CONCLUSION ANDFUTUREWORK

In this paper, we carried out 2.4 GHz IEEE 802.15.4 link measurements to investigate the nLoS link behaviors at three different intersections in the urban area. While the path loss exponents of these nLoS links are always more than 4.2, the links generally have more than 50% PRR when both the transmitting and receiving vehicles are less than 20-30 meters from the intersections. The results suggest that safety applications in VANET should take full advantage of direct nLoS links at intersections, rather than only depending on intermediate nodes to relay the packets. We also proposed a collision avoidance system which can provide drivers/riders information about surrounding vehicles close to the intersection, including their velocities, locations, and gives warnings when collisions are likely to happen; drivers/riders can then react based on the information, just like they have seen these vehicles. Evaluation of the proposed system shows that the system only requires a beacon sending rate of 5 beacons per second to reach a 99% probability of avoiding the collision.

We believe that with the low-complexity and low-cost nature of our proposed system, it is realistic to implement it in current off-the-shelf scooters. Future works include real-world implementation of the proposed system in a testbed of several scooters and evaluation of the system performance in scenarios where scooters are moving at various velocities.

ACKNOWLEDGEMENT

This work is supported by National Science Council, Na- tional Taiwan University, and Intel Corporation under Grants NSC-100-2911-I-002-001 and 10R70501.

REFERENCES

[1] Ministry of Transportation and Communications, Taiwan,

“2011 transport statistics report,” 2011.

[2] IEEE 802.15 Part 15.4-2006 Wireless Medium Access Control (MAC) and Physical Layer (PHY) Specifications for Low-Rate Wireless Personal Area Networks (WPANs), IEEE 802.15 Std.

[3] IEEE Standard for Information technology - Telecom- munications and information exchange between systems - Local and metropolitan area networks - specific re- quirements - Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications

- Amendment 6: Wireless Access in Vehicular Environ- ments, IEEE Std., July 2010.

[4] T. Mangel, M. Michl, O. Klemp, and H. Hartenstein,

“Real-world measurements of non-line-of-sight reception quality for 5.9 GHz IEEE 802.11p at intersections,”

Lecture Notes in Computer Science, vol. 6596, pp. 189–

202, 2011.

[5] T. Mangel, O. Klemp, and H. Hartenstein, “A validated 5.9 GHz non-line-of-sight path-loss and fading model for inter-vehicle communication,” in Proc. IEEE Interna- tional Conference on ITS Telecommunications, October 2011, pp. 75–80.

[6] E. Giordano, R. Frank, G. Pau, and M. Gerla, “CORNER:

a realistic urban propagation model for VANET,” in Proc.

IEEE International Conference on Wireless On-demand Network Systems and Services, February 2010, pp. 57–

60.

[7] C. Sommer, D. Eckhoff, R. German, and F. Dressler, “A computationally inexpensive empirical model of IEEE 802.11p radio shadowing in urban environments,” in Proc. IEEE International Conference on Wireless On- Demand Network Systems and Services, January 2011, pp. 84–90.

[8] Q. Sun, S. Y. Tan, and K. C. Teh, “Analytical formulae for path loss prediction in urban street grid microcellular environments,” IEEE Transactions on Vehicular Technol- ogy, vol. 54, no. 4, pp. 1251–1258, July 2005.

[9] H. Suzuki, H. Murata, and K. Araki, “Performance of inter-vehicle communication technique for intersection collision warning,” in Proc. IEEE Information, Commu- nications and Signal Processing, 2005, pp. 603–606.

[10] “Firefly datasheet,” 2010. [Online]. Available:

http://www.nanork.org/projects/nanork/wiki/FireFly [11] “CC2420: 2.4 GHz IEEE 802.15.4 / ZigBee-ready RF

transceiver, revision swrs041b,” March 2007. [Online].

Available: http://www.ti.com/litv/pdf/swrs041b

[12] “Nano-RK: A wireless sensor networking real- time operating system,” 2005. [Online]. Available:

http://www.nanork.org/projects/nanork/wiki

[13] J. Y. Wong, Theory of Ground Vehicles. Wiley, 2008.

[14] Ministry of Transportation and Communications, Taiwan,

“Table of stopping distance versus speed for vehicles on general roads,” 1977.