Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

IMPERIAL COLLEGE LONDON

CYCSEE
Cyclist Detection Systems
2/27/2012

TEAM
Ali, Zaeed Arif, Bilal Mohammad El-Turk, Hazem Khalab, Ibrahim Liu, Jia Suddle, Ali Tan, Xiao Wu, Jiarui Xue, Zeping

SUPERVISOR: Dr. Timothy G. Constandinou

CYCSEE

Contents
Introduction ................................................................................................................................................. 2 Project Aim .............................................................................................................................................. 2 Project Motivation.................................................................................................................................... 2 Background .............................................................................................................................................. 2 Cycsee technical report ................................................................................................................................ 2 High Level description ...............................................................................................................................3 Low Level Analysis ....................................................................................................................................3 Frequency allocation .............................................................................................................................3 C-Sense (Reader and Alert System) ...................................................................................................... 4 C-Tag.....................................................................................................................................................5 Active tag classification .........................................................................................................................5 Data Processing Subsystem ..................................................................................................................5 Device level analysis and communication ..............................................................................................5 Timing considerations .......................................................................................................................... 6 Cycsee business report ................................................................................................................................. 6 Market Background .................................................................................................................................. 6 Environmental Considerations .................................................................................................................. 7 Nature of market ....................................................................................................................................... 7 Market Research and Commercial Feasibility ........................................................................................... 8 Product Cost Analysis ............................................................................................................................... 8 Financial Projections ................................................................................................................................ 8 Conclusion ................................................................................................................................................... 9 References .................................................................................................................................................. 10 Appendix ..................................................................................................................................................... 11

Note: Throughout the report, Figure App.x refers to a figure in the Appendix, where x represents a number

Page 1

CYCSEE

Introduction
Project Aim
To make cycling a safer and more viable means of transport in congested cities, through the development of a cyclist proximity detector system.

Project Motivation
With an increase in the number of road accidents and roads becoming more congested, cyclist safety has become a real issue for governments and local councils. Every year there are hundreds of cyclist deaths and serious injuries on city roads which could be easily avoided if the right precautions were taken. With the 2012 Olympic Games coming to London, the Mayor and numerous cyclist campaign groups are looking for new systems to provide the safety and security necessary to complete their vision of a green London. To this end, we have proposed a product which aims to create a safer cycling environment Cycsee. This is a concept that utilises Radio Frequency Identification (RFID) technology to alert drivers of the proximity of nearby cyclists.

Background
Over the last decade, cyclist casualties have been steady around the 20,000 per year mark, with a notable increase after 2008. Unlike the decrease in car casualties during the last few years, cyclist casualties have remained constant (refer to Figure App.1). A system is therefore needed to reduce the cyclist casualty rate. Looking more closely at the cyclist casualty figures, we observe that between 1994 2007, the total number of cyclists killed or seriously injured drops, but since 2007 there has been a trend of increasing deaths, making cyclist safety a priority issue.

Figure 1: Cyclist casualty figures over the past decade

Cycsee technical report


Cycsee is an RFID based cyclist detector system which comprises two key components; the C-Tag and the CSense. The C-Tag is an RFID tag which attaches to the bicycle, and its counterpart is the C-Sense which integrates an RFID reader with an alert system, to be fitted onto vehicles. The two components communicate with each other to ultimately inform the driver of a nearby cyclist.

Figure 2: Cycsee consisting of C-Tag and C-Sense

RFID reader

RFID Tag
C-Tag Consists of an RFID tag

Processor Driver Alert


C-Sense Consists of three main blocks

Radio Frequency Identification (RFID) is considered a Wireless Sensor Network (WSN) and is based on radio frequency communication between a transmitter and a detection device. The three core components of an RFID based system are: RFID tag (transmitter) RFID reader (detector) Data processing subsystem In principle, data is stored within the tag and the reader is able to read/write this data when the tag is within its range. The subsystem is then used to manipulate the data in a predefined manner. Page 2

CYCSEE

High Level description


Cycsee requires the C-Sense system to be fitted onto a motor vehicle, and will have two RFID readers located in each of the two rear corners of the vehicle. Each reader will continuously emit RF signals within a specified range. Once a bicycle equipped with a C-Tag (most likely to be incorporated within the bicycle frame) enters the detection zone, it will be registered by the reader and thus its location will be identified. The driver is then alerted of the nearby cyclist. The C-Tag and C-Sense are designed to operate in various environments. The step by step functionality of our system is summarised in the flow diagram below: 1. 2. 3. 4. 5. 6. The reader continuously sends an RF probe wave in search of a tag. When a cyclist enters the detection range of the reader, the tag receives the probe wave Modulates it and emits a response to the reader. The reader receives this wave and demodulates it. The processing unit (microprocessor) operates on the received signal from the reader The alert system then informs the vehicle driver of a nearby cyclist.

6
Key
Reader Tag

Processor

Alert

Low Level Analysis


Frequency allocation One of the major considerations of Cycsee is the selection of an operational frequency. Considering a number of laws and regulations in place in different countries concerning the usage of different frequency bands, coupled with the chance of interference with other communication systems, the 2.4 2.5 GHz frequency band was deemed the most suitable. This is due to three main attributes of this frequency band: - It is a worldwide unlicensed band (ISM) which means there are no governmental regulations on the usage of this band. - It has a bandwidth greater than other frequency spectrums which makes a larger number of communication channels available for its use. - It exhibits good reflection off metal surfaces (which) allows for good propagation in cluttered environments.2

Page 3

CYCSEE Since the 2.4 2.5 GHz band is an unlicensed frequency band; it is widely used and considered to be too crowded. Devices such as radio LANs and Bluetooth Figure 4: Visual representation of Cycsee in action. The bike utilise this band which leads to an increase in the has entered the operating range of the Reader, and the probability of interference from such devices3. The driver is alerted via a flashing bike icon on the dashboard problem of interference can be solved by using a technique called Frequency Hopping Spread Spectrum (FHSS). FHSS works such that readers hop from channel to channel in a pre-assigned, pseudo-random sequence to avoid bumping into each other4. This procedure is particularly appropriate for the 2.4 2.5 GHz band and allows a 100MHz frequency bandwidth to be used. FHSS is suitable for Cycsee because it is a very robust technology, with little influence from noises, reflections, other radio stations or other environment factors.5 An alternative technique, Direct Sequence Spread Spectrum (DSSS) was considered, but discarded because of its lack of robustness. For example, in the worst case, a foreign device such as Bluetooth will only block some hops in an FHSS system while a DSSS system will not work at all under such conditions6. Another significant factor that was considered is the operation of multiple tags at the same time, i.e. when more than one bicycle is present in the vicinity of the vehicle. This situation is known as multi-tag operation and there are anti-collision procedures that can be used to effectively differentiate tags that enter the same reader zone. We have chosen the ALOHA anti-collision algorithm due to its advantages of speed and simplicity.7The basic ALOHA procedure is such that each tag sends its entire ID and then waits for a pseudo-random period of time before broadcasting again. The reader simply receives the IDs depending on chance to ensure that each tag will eventually broadcast during a period when all other tags are quiet.8 The problem with this method, in general, is that it becomes unreliable when eight or more tags9 are present in the vicinity of the reader. The probability of eight or more cyclists entering the range of a vehicle at any one time is almost zero; therefore this disadvantage is nullified for our application and we can benefit from the speed of this procedure, while keeping the solution less complex in comparison to using other possible methods.

C-Sense (Reader and Alert System) The RFID reader interrogates (or reads) an RFID tag when it enters the readers RF portal (reading zone). The RFID reader consists of three main components; an antenna to send and receive RF signals, a microprocessor which is designed to decode the obtained information, and a controller to communicate with the processing subsystem. C-Sense will utilise a circular polarized antenna as this is unaffected by tag orientation10(Figure 4). The advantage of this is that it also has a larger read range and is preferred for an RFID system that operates in the UHF or microwave frequency range11 (Cycsee operates in the microwave region 2.4 GHz). The microprocessor in the reader will decode and process signals received from the tag. The information is then passed on to a controller which is linked to the subsystem through a communication interface, and this controls the alert system. The alert system then informs the driver of the presence of a cyclist, and the side from which the cyclist is approaching.

Page 4

CYCSEE C-Tag For the C-Tag, two possible RFID tag technologies were considered: active and passive. The table below summarises their main differences relevant to our application: Property Read range Power source Required signal strength Set infrastructure cost Cost
Table 1: Passive tag vs Active tag

Passive tag 3m Powered by RF waves High High for fixed readers 0.10 to 3

Active tag Up to 100m Battery attached Low Cheap interrogator (reader) 9 to 30

The table demonstrates that an active tag is more appropriate for Cycsees application. An active tag provides a read range of up to 100m which is an important factor for alerting the driver of the presence of a cyclist and giving them enough time to take any necessary action. In addition, being battery powered means that an active tag has a constant source of energy available for its operation, whereas a passive tag is unreliable in this aspect because it depends on the RF signal strength of the reader. Although the active tag is more costly, the advantages of using it outweigh those of the passive tag. On the basis of this analysis, an active tag is well suited for Cycsee and fulfils the necessary requirements for reliable operation. Active tag classification Active tags can be of two types; wake-up or awake (beacon) systems. The table below highlights the main differences between them12: Wake-up tag system Awake (beacon) tag system Asleep until activated by a coded message from Awake all the time battery power used even the reader when the tag is not being interrogated High data transfer rate Lower data transfer rate and memory size conserves energy Lower component costs
Table 2: Active RFID system classification

Cycsee will be implemented using awake tag systems, as awake tags have proven to be effective in Real Time Locating Systems (RTLS). RTLS are similar to the Cycsee system as they are able to locate a moving object, in our case a cyclist. Awake tag systems also have the advantage of lower component costs and will therefore reduce the cost of the C-Tag. Minimal data will require transfer within the Cycsee system, as such, the awake systems low data transfer rate and memory size are irrelevant. Data Processing Subsystem C-Sense is designed such that it analyses the information received by the two reader antennas on either side of the vehicle. This requires a controller to gather information from the reader and link it to a host processing system. A serial communication interface will provide the host system with the information stored in the reader13. When a cyclist enters the reading zone, tag-reader communication commences and the host system triggers C-Senses alert system to provide a visual and audible warning to the driver. Device level analysis and communication The two tag-reader communication routes that take place in the 2.4 GHz band are Uplink and Downlink. Uplink is from reader to tag and utilises Amplitude Shift Keying (ASK) modulation, whilst Downlink is from tag to reader, and uses backscatter modulation. Figure App.2 provides more detail. The reader within C-Sense continuously emits RF signals. After the C-Tag detects the signal from the reader, it modulates the signal using Downlink modulation. The form of Downlink modulation used is backscatter modulation (also known as On-Off keying). As shown in Figure 5, the modulator (a MOSFET) acts as a switch and is controlled by the output data from the CHIP. When the transistor turns on, it is connected in

Page 5

CYCSEE parallel with the antenna and as a result causes a mismatch in the tag antenna. 14 The output data of the CHIP controls the modulation of the signal and the result is reradiated back to the reader.

Antenna
Figure 5: RFID Active Tag Circuit
15

The receiver on the reader detects this stream of bits and passes it to a Low Noise Amplifier (LNA). This LNA amplifies any weak signals within the stream of bits. The signal is then filtered to remove any noise and oscillations present. A mixer shifts the frequency of the signal and the demodulator uses ASK to represent the binary digits as sinusoidal signals, i.e. recovering the original signal (Figure 5). Finally, this signal is passed on to a Digital Signal Processor the microcontroller part of the reader. This converts the signal into human readable information, which for our application will be both a visual and audible signal alerting the driver of an approaching cyclist. The table below summarise the main characteristics of the C-Tag and C-Sense products. C-Tag Main feature Operating Range Quantity Cost Device suitable for Cycsee Active and Awake 3m 1 per cyclist 9-30 Siemens -MDS U315 C-Sense Integrated antenna (circular polarized), controller and alert system. 3m 2 per vehicle 315-600 plus installation costs Siemens-SLG U9216

Table 3: Main Characteristics of C-Tag and C-Sense

Timing considerations Timing is a vital consideration and it is important to consider the overall RFID system timing in order for Cycsee to work reliably. For example, a MOBY U reader-tag data transfer occurs at a rate of 384kbit/s17. MDS U315 tags have a size of 32 bits and this corresponds to a tag read time of approximately 0.1ms. This means that an RFID reader placed on a vehicle should take approximately a millisecond to read a tag on an approaching cyclist. As mentioned above, Cycsee uses the FHSS method to avoid the problem of interference from other devices. In the 2.4 2.5 GHz frequency band, frequency hopping takes place at a rate of >2.5 hops/sec (Figure App.2). Although this speed depends on the number of foreign devices present in the frequency band, it is a fast enough rate not to cause delays in tag reading for our application. We can conclude that reader-tag operation time is in the order of a few milliseconds. At distances of approximately 3m (our operating range), this communication time will be fast enough for drivers to react to the presence of a cyclist.

Cycsee business report


Market Background
There is currently no widespread implementation of cyclist detection systems on the vehicle market. Products such as blind spot detection systems and See-mi offer similar functionality to Cycsee; however Cycsee holds certain advantages over these applications. Blind spot detection systems aid drivers to avoid hitting other vehicles/objects in their blind spot. These systems consist of two sensing units attached to both sides of Page 6

CYCSEE the back bumper and tend to be costly, varying from 40 to 300, with high-end systems such as that built into the Audi A7 costing over 300.18 See-Mi systems, which also use RFID technology, are currently used in Gren, Denmark. In this system, RFID tags are placed on the bicycle, and the receivers are set at accident-prone crossroads, with drivers being alerted of a nearby cyclist by an additional traffic light. Recently, this system has been incorporated into an urban cycle sharing business venture with Citybike and has brought See-Mi technology to London.19 Cycsee can be viewed as an extension of both these existing technologies. With Cycsee, drivers are alerted of a cyclists presence within larger vicinity. Drivers are informed of a bicycle before it enters their blind spot, allowing adequate time to react. It is therefore a more comprehensive safety system.

Environmental Considerations
Over the past forty years, the number of people who cycle to work has greatly increased, with a particularly steep increase from 2004 onwards (Figure App.3). No doubt this has meant a larger demand for bicycles and is very promising for Cycsees prospects. The more bicycles on the roads, the less the pollution and congestion levels an objective the government has been trying to pursue recently, notably visible when Ken Livingstone introduced the London Congestion charge in 2003. In 2010, the economic cost of road casualties and accidents reported to the police came to an estimated 15 billion. This is a huge cost to the tax-payer, which could be reduced by making cycling safer.
Table 4: Table showing the financial cost of casualties/ accidents, relative to how serious they are. Note that casualty relates to only the human cost, and accident relates to the total cost
20

Looking at the table above, we see that each fatal accident costs the tax payer approximately 1,800,000, an enormous sum. Cycsee will seek to minimise the casualty rates of cyclists by making roads a safer environment for cycling, and therefore reduce the total financial burden on the government and tax-payer. With regards to the environment, pollution is rampant in many major cities across the world and London is no exception. According to Londonair.org.uk, a website set up by the Environmental Research Group at Kings College London, London has one of the highest pollution levels in the UK, and is currently the main area failing to comply with the legally binding limits set by the EU.21 One way to curb this long term issue is through a greener method of commuting cycling. Cycling can also significantly reduce our carbon footprint, via fewer greenhouse gases being emitted, which minimises the contribution to climate change. More cyclists in London imply fewer vehicles on the road, which results in a cleaner and less congested environment.

Nature of market
The commercialisation of Cycsee can be realised through two possible approaches. The first is by seeking government backing in a bid to make it a legal requirement for all bicycles to have a C-Tag attached, and all vehicles to have a C-Sense system fitted. This proposal will require a public bill to be passed, as it will have an effect on all members of the public and thus must be passed through the Houses of Parliament. The second approach is to sell it on the retail market. This can be done by selling C-Tag and C-Sense products to high street outlets, cyclist and vehicle specific stores. An approach such as this will require effective advertising to create the required consumer demand for Cycsee. In addition to stores, the C-Tag could be licenced to bicycle manufacturers and the C-Sense to motor vehicle manufacturers.

Page 7

CYCSEE

Market Research and Commercial Feasibility


We conducted a survey targeted at both cyclists and drivers, with our sample size consisting of 66 cyclists and 57 drivers. From the results, we gathered that cyclists deemed their personal safety on the road as their main concern, with 66% also indicating interest and willingness to purchase a safety device similar to Cycsee. Of the 57 drivers surveyed, 79% acknowledged that cyclists safety was of notable concern to them whilst driving, and 77% agreed that a device such as Cycsee would be beneficial. The most popular alert system amongst drivers (38%) was both flashing lights on the vehicle dashboards with accompanying warning sounds. A promising 36% of our entire sample group showed willingness to buy a safety device such as CTag within the range of 20 - 40. This follows well as our product is estimated to be within this price range.

Main concern regarding cycling


Personal Safety

Willingness to pay
0 - 20

Cycle Security Pollution Lack of cycling infrastructure

20 - 40 40 - 60 60+

Figure 6: Pie charts displaying the results from our survey

Ultimately, the results of the survey showed that there is adequate interest among both cyclists and drivers for a safety device such as Cycsee, reaffirming that the pursuit of this product has great potential.

Product Cost Analysis


The table below shows the estimated costs of both the C-Tag and C-Sense when produced in bulk of 20,000 C-Tag devices and 4,000 C-Sense systems. The cost breakdown is:

C-Tag / cost per unit


Circuit Components Assembly Packaging Branding Storage Transportation Total Cost 2.25 3 1.40 2 0.50 0.85 10

C-Sense / cost per unit


29.50 10 3 2 3 2.50 50

Table 5: Table showing the cost breakdown of C-Tag and C-Sense

Financial Projections
The manufacturing cost and retail price of the C-Tag are estimated at 10 and 15 respectively. Similarly, CSense will cost 50 to manufacture and will retail at 70. The following figures reflect the forecasted profitability of Cycsee in the first five years of production.

Page 8

CYCSEE
Year C-Tag Projected Sales (units) Turnover Gross Profit Overheads Net Profit Total Net Profit
2,000 30,000 10,000 15,000 -5,000

1 CSense
400 28,000 8,000 12,000 -4,000

2 C-Tag
7,000 105,000 35,000 28,000 7,000

3 CSense
2,200 154,000 44,000 35,000 9,000

4 CSense
6,500 455,000 130,000 97,000 33,000

5 CSense
12,000 840,000 240,000 145,000 95,000

C-Tag
26,000 390,000 130,000 91,000 39,000

C-Tag
50,000 750,000 250,000 165,000 85,000

C-Tag
130,000 1,950,000 650,000 300,000 350,000

C-Sense
25,000 1,750,000 500,000 280,000 220,000

-9,000

16,000

72,000

180,000

570,000

Table 6: Net Profit for both C-Tag and C-Sense in the first five years of production.

SWOT Analysis

First Five Year Profit Projections of Cycsee


600,000 500,000 400,000
C-Tag Net Profit C-Sense Net Profit Total Net Profit

Amount

300,000 200,000 100,000 0 (100,000) 0

2 Year

Figure 7: Graph showing the expected profit trends for Cycsee in the first five years

Conclusion

Figure 8: SWOT Analysis identifying internal and external factors that affect our objectives

Based on the technical and market analysis conducted, it is evident that there is demand for a cyclist safety device, the criteria of which Cycsee will satisfy. No such safety system is currently available across London, and so in line with the Mayors future transport strategy and goal of making a greener London, Cycsee has a very promising potential in this otherwise undeveloped market. In conjunction with the current product proposal being put forward, Cycsee can be expanded to resolve many other issues cyclists face, other than their personal safety. For example, C-Sense hubs can be set up across the capital to mark popular cyclist routes and accident hotspots, in a similar manner to the See-Mi system in Denmark. Analysing statistics has shown us that there has been a slight increase in cyclist casualties over the past few years, which the successful implementation of Cycsee should see the reduction of, and thereby reduce the human costs as well as the financial costs to the government. This should act to encourage commuters to choose cycling over other means of transport, thus reducing both pollution levels and congestion. Cycsee has the potential to revolutionise road transport and provide cyclists with an almost risk-free cycling environment.

Page 9

CYCSEE

References
[1]. DfT STATS19. Reported pedal Cyclist casualties GB 1979-2010. [Online]. Department for Transport Statistics. Available from : http://www.dft.gov.uk/statistics/tables/ras30065/ [Accessed 29th December 2011] [2] Lehpamer, Harvey. RFID Design Principles. United States of America. Artech House. 2008. pp. 119. [3] ibid pp. 120. [4] ibid. [5] M. Schwartz, Sorin. FHSS vs. DSSS in BWA and WLAN. [Online] Sorin M.Schwartz Seminars. Available from: http://sorinschwartz.com/white_papers/fhvsds.pdf [Acessed 10th January 2012] pp. 3 [6] ibid pp. 10 [7] Glover,Bill, Bhatt, Himanshu. RFID Essentials. [Online] O'Reilly Media, Inc. 2006. Available from: http://books.google.co.uk/books?id=K2gdK21RVEC&printsec=frontcover&source=gbs_ge_summary_r&cad=0#v=onepage&q&f=fa lse [Acessed 5th January 2012] - pp. 88 [8] ibid. [9] ibid. [10] Lahiri, Sandip. RFID Sourcebook. [Online] Pearson plc, 2006. Available from: fromhttp://my.safaribooksonline.com/book/hardware/rfid/0131851373/firstchapter#X2ludGVybmFsX0ZsYXNoUmVhZGVyP3htbGl kPTAxMzE4NTEzNzMvdmlp [Acessed 15th January 2012] - pp. 36 [11] Lehpamer, Harvey. RFID Design Principles. United States of America. Artech House. 2008. pp. 120. [12] ibid. pp. 162. [13] Lahiri, Sandip. RFID Sourcebook. [Online] Pearson plc, 2006. Available from: fromhttp://my.safaribooksonline.com/book/hardware/rfid/0131851373/firstchapter#X2ludGVybmFsX0ZsYXNoUmVhZGVyP3htbGl kPTAxMzE4NTEzNzMvdmlp [Acessed 15th January 2012] - pp. 36 [14] Lehpamer, Harvey. RFID Design Principles. United States of America. Artech House. 2008. pp. 169 [15] ibid. pp. 63 [16] Siemens AG 2011, RFID systems for the microwave frequency range [Online] Siemens Technologies. Available from: http://www.automation.siemens.com/mcms/infocenter/dokumentencenter/sc/fs/Documentsu20Catalogs/E86060-K8310-A101-A77600-ID10_2011_en_kap_4.pdf [Accessed 7th January 2012] [17] Manuals MOBY U, Manual on configuration, installation and service of the identification system. Edition: May 2004 - [Online] - Available from: http://support.automation.siemens.com/WW/llisapi.dll?func=cslib.csinfo&lang=en&objid=6GT25010BA00&objaction=csviewmlfbbeitraege&subtype=133300&caller=view [Accessed 27th January 2012] [18] Russin, Nina. 2012 Audi A7 Sedan. [Online]. Available from: http://carspondent.com/2012-audi-a7-sedan/ [Accessed 27th January 2012] [19] Lynch, Ken. Can You See Mi Now? [Online]. Available from: http://rfid.thingmagic.com/rfid-blog/bid/50972/Can-You-See-MiNow [Accessed 20th January 2012] [20] Kilbey, Pat. Reported Road Casualties Great Britain: 2010 Annual Report. [Online]. Transport Statistics: DfT. Available from: http://assets.dft.gov.uk/statistics/releases/road-accidents-and-safety-annual-report-2010/rrcgb2010-complete.pdf [Accessed 15th December 2011] pp.47. [21] London Air. Is air pollution worse in London? [Online]. Available from: http://www.londonair.org.uk/LondonAir/guide/London.aspx [Accessed 18th January 2012] [22] Kilbey, Pat. Reported Road Casualties in Great Britain: Quarterly Provisional Estimates Q2 2011. Department of Transport. 2011 [23] Edgar Sanchez-Sinencio, RF Communication Circuits and Systems Lecture 4 .[Online] Texas A&M University Available from: http://amesp02.tamu.edu/~sanchez/RFID_665.pdf [Accessed 10th January 2012] pp.31 [24] London Transport Data. Long-run trend in commuting into central London [Online]. Available from: http://londontransportdata.wordpress.com/category/type/data/ [Accessed 20th January 2012]

Page 10

CYCSEE

Appendix

Figure App.1: Graph showing road casualties by road users, from Quarter 2 2002 22 Quarter 2 2011

Parameter Reader-Tag communication (Uplink) Tag-Reader communication (Downlink) Operating channels Bandwidth FHSS
Figure App.2 2.4GHz standard parameters and values
23

Value ASK modulation Backscatter modulation 79 channels from 2422.5 MHz to 2461.5MHz Maximum of 0.5MHz Hop rate > 2.5 hops/sec

Figure App.3: Graph showing an increase in cycling

24

Page 11

You might also like