safety improvements using microwave networks Pavel Rozsival, Jan Pidanic, Petr Dolezel, Pavel Bezousek Department of Electrical engineering, University of Pardubice, Pardubice, Czech Republic [email protected]

Abstract - Article describes use of radio network using 2 Possible Solution 2.45GHz (ISM) frequencies for safety management and Our scenario uses radio nod on every train car used on logistic in train transportation. Possible uses and expected selected tracks, around these tracks are radio base stations properties of low cost network based on radio nod equipped placed on important parts of . Nods should be train wagons are described. Design of such radio network universally reconfigurable to act in different ways. based on NRF24L01 radios is outlined. This paper also describes testing scenarios for estimated parameter proofing. This system can work in more ways, to increase safety or At the end of the paper are presented results from testing comfort of train use. At first, if every nod contains specific ID dependency of movement on reliability of communication and code, it can be used for identification. In this case, it can help first results from real scenario tests. Data shows that concept with train composing. Now the worker have to go to vehicle of placing low cost radios on wagons can be easily used in read the number, check and type in, other one is selecting various situations like radio identification even in high speeds manually and composing by the list. RFID based train cars between train and ID reader. can make this partially autonomous. It can help to detect

Keywords: train; safety; RFID; NRF24l01 presence of in parking places, train stations, maintains

stations if equipped with ID readers. Also tracks equipped with readers can detect presence of every car. System 1 Introduction equipped with readers among tracks can detect differences in Railway transportation is historically one of the most safe cars passed through check points and detect train integrity and reliable way of transport people and goods [1] [2].With loss. System can be also used for positioning of trains with increased traffic on railroads a lot of safety issues were solved dangerous or valuable freight or even freight itself. and are still in solve. These days safety management in train transportation is one of the most developed, but still dealing NODs don’t have to act only as ID radio tag, but can also with issues. Some errors are unpredictable as human being contain sensors to measure conditions for fault prevention, or and mostly caused by human [3]. Most of the mistakes are damaged goods complains and much more. Radio tags can easy to prevent or fix, like entering wrong track or direction. also create networks and detect changes in neighbors. But some of them are still uncovered by any safety system. Networks can be configured by higher system, or can be made as self-configurable, but because of possible presence One of the unsolved problems is train integrity detection of nods that are not part of the train or low possibility to for example. Usually most of automatic safety features is detect faulty NOD, it’s better to use higher system to connected to , but if the train loses some wagons, configure train car network. This network can easily detect it’s hard to detect them on tracks, or especially in freight neighbor loss and create integrity warning. [4] trains to even notice missing car. Another problem is identification of trains or train wagons, most of tracks and Because of huge amount of train cars, radio nods should safety electronics are able to detect train on track, but only be as cheap as possible. Nods have to offer long live span. few are equipped for reading ID of train, and again ID is Because of train speed varying from 0-160 km/h (in the placed on locomotive only. Czech Republic) system might be able to work up to 200 If all train vehicles including both and km/h. Also tags will be exposed to weather conditions, so wagons are equipped with programmable radio NOD it can they should withstand temperatures from -20 to 60°C, high help to solve, or be a part of solution of most of these issues, humidity or running water contact. and secondary help with other non-safety related demands. 3 System design microcontrollers. The Nordic nRF24L01+ integrates a complete 2.4GHz RF transceiver, RF synthesizer, and 3.1 Radio network baseband logic including the Enhanced ShockBurst™ hardware protocol accelerator supporting a high-speed SPI Radio network is composed from nods and base stations. interface for the application controller. No external loop filter, In our testing scenario, every carriage is equipped with 2 resonators, or VCO varactor diodes are required, only a low nods, one nod on every side of train car preventing radio cost ±60ppm crystal, matching circuitry, and antenna. The shadow creating and increasing reliability by this doubling. nRF24L01+ comes in a compact 20-pin 4 x 4mm QFN Radios must be placed on the outer side of wagon due package [5]. metallic (short wave radio proof) walls of trains. Side of under frame was chosen. Place is covered from the top, hidden from eyes and doesn’t affect the profile of the train.

Figure 2: Radio chip connection diagram Microprocessor selection is not as important as radio chip.

Low power, low voltage and low cost are the only demands. Figure 1: Tag/Reader placement Processor from Atmel AVR family was chosen, namely ATmega48PV. ATmega48 is the cheapest available processor Rails in testing area are equipped with base stations from Atmel, it is important to choose PV suffixed parts. P is collecting data on passing trains. Most of stations are suffix for processors with picoPower technology with very equipped with GSM modem for data retransmit to central low power consumption sleep modes. V stays for low voltage. database for later processing. Processors with V in name are able to run from 1.8-5.5 V compared to others that works in range from 2.7 to 5.5 volts. 3.2 Tag This processor offers plenty of peripheries that can be used for extending functionality of application. In our case only SPI Tag is radio NOD basically programmed to transmit port is used to connect radio chip and UART for debugging or unique ID for wagon identification. connecting when the nod is used as transceiver.

Nod as itself is consisted of radio chip, low power microprocessor, battery and few discrete components.

Selection of radio components highly affects properties of whole system. Trying to keep hardware part as simple as possible, we have to use radio chip with high level of integration of all major radio parts and low support components count. As a compromise between simplicity and efficiency NRF24L01 from Nordic semiconductor was chosen.

The Nordic NRF24L01+ is a highly integrated, ultra-low power (ULP) 2Mbps RF transceiver IC for the 2.4GHz ISM (Industrial, Scientific and Medical) band. With peak RX/TX currents lower than 14mA, a sub μA power down mode, advanced power management, and a 1.9 to 3.6V supply range, the nRF24L01+ provides a true ULP solution enabling months Figure 3: Microprocessor connection to years of battery life from coin cell or AA/AAA batteries. Chip antenna solution was used for tags as space saving The Enhanced ShockBurst™ hardware protocol accelerator solution. For special purposes was developed nod with SMA offloads time critical protocol functions from the application connector for external antenna connection. microcontroller enabling the implementation of advanced and robust wireless connectivity with low cost 3rd-party Battery selection must be compromise between size and capacity. Lifespan of tag should be at least 1 year. First iteration used CR2450 battery offering theoretical lifespan over 2 years. Testing in climatic box showed that battery based on Li-MnO2 suffers from extreme capacity los under low temperatures. In other iterations CR14250 battery with Li-SOCl2 mechanism were used. This battery is only slightly affected by temperature and offers 3 times higher capacity compared to CR2450, with only small size adding (tag with this battery is 5 mm higher). Figure 7: Final tag For outside condition deployment, tag is fitted in Hammond1551 flanged or non-flanged 20x35x60 mm box. 3.3 Base station Base stations are special kind of NOD, usually consisted from tag electronics acting as transceiver and higher computer that offers better connectivity (serial, Ethernet…) or functionality (database, data logging, data analysis…). We are using 3 kinds of base stations. First one is mobile base station consisted from tag with external antenna connector and GSM modem for data harvesting and retransmitting in situation where other data network connection is not possible.

Figure 4: Nods

Figure 5: Boxed Nods Due to high parasitic properties of component used in this solution, range of radio showed high variation in distance. Figure 8: Reader with GSM modem Tag design was improved to 0402 sized components, increasing stability of design and decreasing size of board. Second one is tag connected to 3.5” industrial PC. PC based computer PEB2737 offers all possible connections and programing versatility. Low cost demand for base stations is no longer needed. Amount of base stations needed compared to tags is negligible.

Figure 6: Board size comparsion Size of the board was decreased to 15x45 mm with all components on one side and more peripherals taken out. Design upgrading to 0402 passive components showed huge improvement in stability and radio performance. Smaller PCB with components on one side decreased costs and production time. Figure 9: Base station with embeded PC Table 1:Hits vs. expected ratio 3.4 Principle of operation reading/distance 5m 10m 30m 50m 60m NOD programmed as radio ID tag is acting like active tag 4 100% 100% 100% 100% 96% RFID, which means it transmits unique ID. This can be done tag 5 13% 0% 0% 0% 0% by two ways, transmitting ID every fixed moment, or on tag 7 100% 100% 97% 93% 67% demand from reader. First option was selected for testing tag 11 100% 98% 100% 96% 74% purposes. In this case whole tag can stay in sleep state with tag 14 100% 93% 91% 91% 87% very low power consumption for most of the time. As a Table shows that most of NODs are able to communicate compromise cycle of 1 sec was chosen. It means that on distance greater than 50 m. Tag 5 vas found faulty and communication usually takes 1ms with 20-25 mA discarded. consumption and rest of second stays in sleep mode with around 6A consumption. This enables tag lifespan for years. Last test was relative speed test. This test was performed Communication is addressed and packet oriented with CRC at the Hradec Kralove airport, where the nod was placed in enabled, using 4 bytes of data as ID. the middle of tarmac and base station in the car. Car was passing in different speeds up to 180 km/h. Due the weather Base stations acts as RFID readers, they are trying to condition of that day, it was impossible safely reach 200 collect ID from all passing trains and put the data into km/h. Positive ID count readings was logged. database. At the very beginning of experiment, base stations were made as self-powering, and hanged on power electricity • 140 km/h 4 hits wires holding posts. Base stations are always in active mode; • 90 km/h 3 hits (returning car) this makes them more current hungry. Short time between • 140 km/h 4 hits battery change and attack of vandals made us move them in • 120 km/h 2 hits (returning car) backup boxes that are along the rails, safely closed and • 140 km/h 3 hits providing power supply. All stations were equipped with • 150 km/h 2 hits (returning car) GSM modem, because of lack of other networking option. • 160 km/h 3 hits • 180 km/h 2 hits (returning car) • 160 km/h 2 hits 4 Pre deployment testing • 180 km/h 2 hits (returning car) Before the system was deployed on railway, basic parameter testing was done to determine behavior and some parameter tuning. Car was moving in both directions, and the position of equipment was mirrored on return, this can change radio Measured parameters are power consumption of tags, wave propagation. Setup of this test was reversed compared communication distance (tag to base station), and effect of to real scenario, tag was placed outside and base station was relative speed. placed in car. This setup was chosen because of heavy rain falls. Weather condition also made impossible to safely drive Power consumption was measured with scope meter car in higher speeds and keep this speed for 100 m around the connected to resistor placed between battery and tag. tag. Measured consumption was varying around predicted values, around 6-7 A for 998 ms and 20-30 mA for 1 ms, this means This test proved presumption of tag requirements and setting. 50 m+ range, 1 ID per second is compromise setting average power consumption is max 37 A. for still reliable setup. If higher ratio of possible reading is needed to increase probability of safety ID, repeating period Maximum communication distance is also important. can be shortened but for the price of battery live drop, almost During radio identification on tracks, train can be passing at all battery consumption is made when on air, this means full speed, Czech limit is 160 km/h but with some reserve we doubling communication will cut battery live to half. This wanted to guarantee up to 200 km/h. While the car is passing experiment also proved minimal impact of Doppler frequency in speed of 200 km/h it will move 56 m every second. If we shift calculated to 133-570 Hz [6]. have reader with radial pattern antenna and tag communicating every 1s. We need reading distance at least 50 m to guarantee at least 1-2 readings by covering area of 100 5 Preliminary testing results m. For this test 5 Tags were picked and placed in open field For preliminary testing one train car was marked with tags with no obstacles in 600 m radius. Base station with ¼ wave and two readers were deployed. Two tags were used, one tag antenna was moved until communication was lost. 1 way per side of the train car. traffic from tags, 1 transmit per second was used, table 1 shows received vs. expected amount of data.

Figure 10: Tag placement Base stations were placed on both sides of tracks in different locations. Reader 1 was mounted 2 m from tracks and battery operated; antenna was approximately in same high as tag on wagon. Reader 2 was mounted in rail house with Figure 13: Position of reader 2 electronics close to railroad crossing and this reader was live power operated for the price of 11 m distance through plastic composite wall, also antenna was 1-2 m higher than ideal position.

Figure 14: Reader situation

Figure 11: Position of reader 1

Figure 15: Reader 2 placed Table 2: Hits per passes of the train

Tag no. reader 1 reader 2 passes Near side ratio Far side ratio 13 17 10 17 100,00% 58,80% 10 13 15 17 88,20% 76,50%

You can see impact (table 2) of reader placement on system properties, from later signal propagation measurement in place, reliability can be increased by using antenna with shaped pattern. You can also see that it’s hard to communicate with module covered by body of the train (far Figure 12: Reader 1 placed side reading). Special pattern antenna based on collected data is under development.

Figure 17: Reader 1 placement Second reader was placed close to Uhersko in security booth only 3 meters from tracks. Figure 16: Train with tags (marks on sides); coverage by radial pattern antenna (circle) and modified pattern antenna (eliptic) antenna and directional antena (U shape); antenna position is in the middle of the circle. 6 Full scale test Previous testing showed proof of concept of our design. Some minor bugs were fixed and tag design was upgraded. Before full test placement speed test was repeated. This time with multiple tags placed and condition good enough to reach speed of 205 km per hour. At least 2 positive ID logged were taken as positive identification.

• 190 km/h 4/4 logged • 185 km/h 4/4 logged • 188 km/h 3/4 logged (possible shadowing) Figure 18: Reader 2 placement • 205 km/h 4/4 logged • 205 km/h 4/4 logged

After this recheck preparation of full scale long time testing took place. First goal was to find appropriate tracks and positions for reader placement with heavier train traffic. After discussion main corridor was chosen and places for readers were found. Because of long time testing expectation, places with power supply and GSM signal coverage was needed.

First reader was placed in Pardubice/Slovany in train security booth 8 meters from tracks. Trains are passing in almost full speed on two parallel tracks.

Figure 19: Reader 2 antenna placement Second and third reader uses concept of directional antenna pointing in sharp angle with tracks. Panel antenna with 60 degrees pattern was used. safety improvement, but mostly as a support not as the only solution. But for the price less than 10 USD for prototype per radio nod (4-5 USD in larger volumes) it’s still huge improvement compared to price paid.

Long time testing should last for one year to show efficiency, reliability and security of this approach.

Questionable is radio band selection 2.4 GHz offers fast data transfer, leading to short time to ID, sub GHz offers higher range with lower power. Interesting option is 5GHz band but with lack of selection of integrated radios. Ultra Figure 20: Reader situation wide selection of various integrated radios in 2.4 GHz kept us Third Reader was placed in Sedlistka close to the railway in this band. station in train security and driving house. It’s the only concrete house used. Acknowledgment

The research was supported by the postdoctoral project Strengthening of Research and Development Teams at the University of Pardubice No. CZ.1.07./2.3.00/30.0021 and by the Czech Ministry of Industry and Trade project No. FR- TI1/084. At this point I want to thanks for collaboration and help to DKV (Railway Car Depot) Pardubice with tag placement and SZDC (Railway Infrastructure Administration) Pardubice with reader placement.

8 References

[1] Wikipedia, "Aviation safety," 10 april 2013. [Online]. Available: http://en.wikipedia.org/wiki/Aviation_safety. [Accessed 2013 april 2013]. Figure 21: Reader 3 situation [2] R. Ford, "INFORMED SOURCES," 2010. [Online]. One reader was left on track used in preliminary testing to Available: cover trains that are not returning directly but circling around. http://dspace.dial.pipex.com/town/square/ca14/ALYCIDO Position of antenna was rearranged in correct high. N%20RAIL/INFORMED%20SOURCES%20ARCHIVE/ INF%20SRCS%202000/Informed%20Sources%2010%20 In this time tags are being placed on cars, and only partial 2000.htm. [Accessed 10 april 2013]. amount of them is currently in process. But until now all of [3] "Lists of rail ," 10 april 2013. [Online]. them are working perfectly. Even in accidental situation when Available: tags were transported by super city train travelling mostly in http://en.wikipedia.org/wiki/Lists_of_rail_accidents. full speed around 160 km/h (speed showed by infotainment [Accessed 10 april 2013]. system in train) through installed area 12, 9 and 13 from 20 were identified when passing around reader 1, 2, 3. Tags were [4] H. Scholten, R. Westenberg and M. Schoemaker, placed in static dissipative bag in luggage inside of the train. "Sensing Train Integrity," in Sensors, 2009 IEEE, Chrischurch, New Zeland, 2009. Collected data are stored on server and displayed on web server application. Application is accessible online on http://vlaky.eparo.cz. In this time it is too soon to create statistic report but at least we can say it looks like it works.

7 Conclusion This article shows the possibilities of radio network used for safety and logistics management. Preliminary testing shows that covering trains by radio network can be used as a