Drive-by-Wireless with the eCar Demonstrator

1

Drive-by-Wireless with the eCar DemonstratorHauke Stahle, Kai Huang, Alois Knoll

Abstract—Drive-by-wire technology has been gradually adopted by the car companies in recent years to integrate active assistancesystems in vehicles to increase comfort and safety. To push the technologies for the electronic control systems of vehicles to extreme,we investigate the so-called drive-by-wireless, i.e., using wireless network to control steering, braking, accelerating, and other functionswithin an automobile. We use commercial off-the-shelf ZigBee MSP-EXP430F5438 Development Board for wireless communicationand demonstrate our drive-by-wireless prototype on a 4-wheel steering/drive electric vehicle.

Keywords—Drive-by-Wireless, Electric Vehicles, Safety, Latency

F

1 INTRODUCTION

In recent years, drive-by-wire technology has beengradually adopted by the car companies to integrateactive assistance systems in vehicles to increase comfortand safety [1]. Furthermore, complete new vehicleconcepts are possible caused by the flexibility of drive-by-wire technology. In general, those techniques replacethe mechanical and hydraulic connections between thedriver and the associated vehicle actuators with elec-tronic communication systems. These systems transmitelectronic messages to direct a vehicle component basedon the action taken by the driver of the vehicle, e.g.,turning a steering wheel, depressing a brake pedal, ordepressing an accelerator pedal. The advantage of drive-by-wire is that safety can be improved by providingcomputer controlled intervention of the driver’s com-mands with systems such as crash avoidance, brakeassistance, and lane assist systems.

To push further the technologies for the electroniccontrol systems of vehicles, we investigate drive-by-wireless, i.e., using wireless network, rather than elec-tronic cables, to control steering, braking, accelerating,and other functions within an automobile. The advan-tage of using wireless is multifold:

• Eliminating wired connections between sensors,actuators, and control modules so as to, in turn,eliminate associated design, installation and main-tenance costs of those wired connections [2]. Forinstance, the possibility of the malfunction of oneor more wires may be eliminated through theuse of a wireless network and the complexity ofmaintenance, problem solving, and repair may bereduced when wiring has been eliminated as apossible source of malfunction. In addition, thereadiness of commercial off-the-shelf communica-tions modules for wireless protocols, such as GSM,3G, LTE, ZigBee, and WiFi, are available, which also

• Hauke Stahle, Kai Huang, Alois Knoll are with the Department ofInformatics, Technische Universitat Munchen, Germany.E-mail: [email protected]

results in simple design and lowered installationand maintenance costs.

• In some wired networks, all modules may sharethe same communication media. The capacity ofthe wired network may become congested anddevelop unacceptable latency. A wireless networkfor controlling vehicle functions may reduce thecapacity limits associated with a wired network,e.g., using different frequency domains for differenttypes of communication.

• A wireless network also may increase the flexibilityof design options because sensor, actuators, andcontrol modules may be located without regard forwiring requirements [3]. The installation of sensormodules and control modules may be easier whenwires need not to be installed between a sensormodule and a control module. This flexibility isimportant for designers and manufacturers, i.e., thesimplification of assembly and the possibility of newmodular design concepts.

• The integration of more and more safety functionsincreases the demand of inter-domain data trans-mission within the vehicle. The different domainsof a car are connected nowadays via a centralgateway which has to handle the cross-traffic ofseveral bus systems with different technologies. Dueto the availability of different channels for wirelesscommunication, a node is not fixed to a specificdomain anymore and can dynamically be groupedwith senders and receivers. This eliminates thebottleneck of a central gateway within a vehicle.

Although drive-by-wireless conceptually provides lotsof advantages, it also faces practical challenges. Twomajor problems are communication latency and safety.The latency issue comes from the additional delayswhich the wireless network induce due to the usedprotocol and propagation delay. The safety issue stemsfrom the environment effects and interference from othercommunicating entities by which the wireless signalcould be easily perturbed.

In this work, we mainly focus on the aforementioned

2

Fig. 1: fortiss eCar demonstrator - an experimentalplatform for innovative car architectures.

two challenges. A wireless deployment environmentis setup based on a drive-by-wire four-wheel steer-ing/drive electric vehicle (Fig. 1). To verify the con-cept of drive-by-wireless, ZigBee protocol is currentlychosen to steer, accelerate, and decelerate one of thetwo axels. Commercial off-the-shelf MSP-EXP430F5438Development Boards are used to integrate the ZigBeecommunication into the existing ICT infrastructure [4] ofthe eCar. To cope with interference between different in-vehicle ZigBee modules, a time division multiple accessschedule is designed to provide temporal isolation. Totackle environmental effects and malicious hijacking, thebeams of the eCar are used as a waveguide for theZigBee communication. Our experimental results showthat the communication latency and safety issues canbe nicely tackled and the concept of drive-by-wireless ispractical feasible on our vehicle.

The rest of this abstract is organized as follows: A shortintroduction for the fortiss eCar is presented in Section 2.Section 3 and Section 4 details our communicationprotocol and safety setup. Section 5 concludes theabstract.

2 ECAR

The complete drive power of the eCar [5] is 8kW. Each ofthe in-wheel engines has 2kW and a maximum torque of160Nm. The eCar is controlled with a sidestick connectedto the human-machine interface unit. The current stateof the evaluation platform is presented on a 10-inchtouchscreen, which can also be used as an input device,i.e., to change between different driving modes. Theoutline of the eCar is approximately 2.25 x 1.25 x 1.75m (L x W x H) and its weight is about 600 kg. The eCaris constructed to carry one passenger.

The ICT infrastructure of the eCar, shown in Fig. 2,consists of four nodes: Three Texas Instrument LM3S8962evaluation boards for the control functions of the eCarand one Intel Atom computer for the management ofthe human-machine interface. The evaluation boards areinterconnected via Ethernet, running a real-time proto-col. FreeRTOS is used as real-time operating system forthe evaluation boards to manage the communication and

Fig. 2: Information and communication architecture ofthe eCar with wired and wireless data transmissions.

applications timings. The computer for the managementof the human-machine interface runs Linux and showsan interface based on the Qt framework.

The connection between the evaluation boards andthe motor and steering controllers is realized via CAN-Bus (Controller Area Network), running the OpenCANprotocol. The central evaluation board and the human-machine interface are connected via a serial link.

3 COMMUNICATION LATENCY

The goal of our deployment setup is to replace allthe Ethernet cables in the eCar with ZigBee wirelesscommunication and achieving a comparable perfomanceto the wired case.

In our first experiment, we replaced the Ethernet cableto the rear axle with a wireless connection while leavingthe setup of the front axle untouched, see Fig. 2. Thisallows us to compare the perfomance of the two differentapproaches.

The wireless nodes are connected to the centralcontroller and rear axle controller via a serial connection.The connection between the wireless host boards and thedaughter boards with the actual transceiver is realizedvia Serial Peripheral Interface (SPI).

While the controllers of the front axle and thecentral system are synchronized via the flexible time-triggered ethernet (FTTE) protocol, the controllers ofthe central system and the rear axle are not, as thesynchronization data is not forwarded via the wirelesslink. Nevertheless, the wireless system itself is runninga beacon-based communication schedule to guaranteecollision-free communication.

The schedule for the wireless communication is basedon a beaconing approach. Although not necessary for thefirst experimental setup, the schedule was designed toserve a maximum of five nodes. The master node sendsa beacon every 30ms and the slave nodes respond afterreception of the beacon and an individual delay, see alsoFig. 2. The transmission time of one packet is 4ms andafter a silence phase of 2ms for the processing ot theincoming data, the next node will send its data. Eachpacket has a size of 125bytes with 100bytes available

3

Fig. 3: The used time divison schedule for wireless datatransmission.

for payload and is transmitted with 250kbit/s over thewireless media. In case of a corrupted transmission, thepacket is lost and not re-transmitted.

Fig. 4 shows the latency caused by other componentsof the system in the complete chain from the sidestick tothe actuator. The communication between the hardwareunits is realized via buffers. Incoming data is stored ina buffer until a periodic task processes this data andwrites the outgoing data into an output buffer. The datastays in the output buffer until a sending task picksit up and sends it via wire or wireless. The period oftasks processing incoming data is labeled ”d in” in Fig. 4while the period of tasks transmitting the output buffersare labeled ”d out”. A value of 0 indicates that the datais processed event-based. The latency of the controllersof the actuators is unknown, thus no values are given.Note that this representation is not sufficient to calculatea tight end-to-end delay as the interaction of differenttasks within one controller are not considered.

First tests with this setup proved, that the achievedlatency and transmission rate is suitable to control therear axle of the eCar. In the near future this setup willbe extended to also allow a wireless transmission to thefront axle and the battery system making the ethernetbased part obsolete. Also, we want to reduce the delaywithin the whole system to make it react faster to thedriver’s inputs.

As the slave nodes only react on the beaconing signal,a loss of the master node causes the complete wirelessnetwork to stop. Therefore, we want to extend ourprotocol to support redundancy via multiple transceiversand to be able to detect the absence of a beacon. Inaddition, we are developing a formal model for oursetup, trying to provide a theoretical bound for the worstcase end-to-end latency of our wireless communication.

4 SAFETYThe wireless transmission inside a vehicle is subject toelectromagnetic interference. This can either be causedby other devices inside the vehicle, devices outside thevehicle or even caused by an external attack with theaim of a directed manipulation of the infrastracture.

The effect can reach from the degration of thecommunication perfomance up to a complete loss of the

Fig. 5: Utilizing compartmens of the vehicle as waveg-uides.

(a) Antenna extension (b) Waveguide (c) Final setup

Fig. 6: Using the beam of the eCar as a waveguide.As there is an on-board antenna, a simplified insulationpack is built to shield signals from the on-board antenna.

communication ability. To prevent a malfunction of thesystem, we propose to use the internal compartmentsof the vehicle as a waveguide. This will hinder externalsignals to interfere with the wireless intra-vehicle com-munication. This concept is shown in Fig. 5.

With the availability of such a compartment that canact as a waveguide inside the vehicle, additional nodescan easily be integrated into the system by installing theantenna in the waveguide.

For our first experiments, we used the beams of theeCar as waveguides for the ZigBee communication toprevent malicious hijacking and to reduce environmentaleffects.

The experimental scenarios are shown in Tab. 1.Basically, we record data acceptance rate betweentwo ZibBee modules, under different combinations of:jammer on/off, antenna in/out of the beam, and endof the beam covered/open. The impacts of the distanceand allocation of the jammer to the signal source aremeasured as well. In Fig. 7, j-c-2 represents that thejammer is two meters away from the signal source withthe end of the beam covered with insulation materials(Here, u represents an open end of the beam). j-c-10-arepresents a 10cm distance between the jammer andthe signal source. The postfix -a and -b indicateddifferent allocations between the jammer and source. Asa signal jammer, we used another ZigBee node that sendsrandom data continously on the same channel.

From Tab. 1, we can clearly see that the beam can

4

Fig. 4: Timings for the transmission of commands from the sidestick to the eCorner. d in and d out refer to theperiodicity of the tasks processing the incoming and outgoing buffers, respectively.

TABLE 1: Experiment scenarios and results. The numberindicate the acceptance rate. One antenna was alwaysleft inside the beam with an insulated beam end. Theposition of the other one was changed according to thetable.

jammer off jammer onantenna covered 100% see Fig. 7

inside beam not covered 100% see Fig. 7antenna covered 0% 0%

outside beam not covered 100% N/A

be used as an insulation media to protect the wirelesssignal from external interference for the scenario, wherethe open end of the beam is covered while the antennaof one device is outside the beam. The packet loss forthis scenario is 100 percent.

The results shown in Fig. 7 are more interesting.In general, the jammer will affect the data acceptancerate. With the covering of the beam, the acceptancerate improved, which can be seen in the cases j-c-2,j-c-10-a, and j-c-10-b. However, the data accep-tance rate cannot be recovered to 100% for all caseswith a closed-end beam. The reason is, that there is anonboard antenna on the wireless module which cannotbe put into the beam. Although we have made cover forthe onboard antenna in order to insulate it, the signalis still affected. Another phenomenon is that the relativeposition between the jammer and signal source affectsthe acceptance rate as well, which is subject of furtherinvestigation.

5 CONCLUSION

We have proved with our experiments, that the conceptof drive-by-wireless is feasible to control the eCardemonstrator. We tackled two of the main challengestowards a reliable drive-by-wireless concept: Latencyissues by using a time-divison protocol for the wirelesscommunication and safety issues by utilizing the beamsof our demonstrator as waveguides to shield the wirelesssignals from external interference. The results showthat the latency is low enough to control the eCar

Fig. 7: Detailed results for the acceptance rates indifferent scenarios. j=jammer active; c=end of beamcovered; u=end of beam uncovered; 2=distance of 2m;10=distance of 10cm; a,b=different positions of wirelessnodes; the results are given as percentages.

and the concept of waveguides greatly improves thepacket acceptance rates between the communicationparticipants in the case of external interference.

In future work, we will investigate in an improvementof the shielding, a lowering of the latency to increaseresponsiveness and a comparison of suitable protocolsfor the in-vehicle use of wireless data transmission.

REFERENCES[1] fortiss GmbH, “The software car: Information and communication

technology (ict) as an engine for the electromobility of the future,”Tech. Rep., March 2011.

[2] M. Ahmed, C. Saraydar, T. ElBatt, J. Yin, T. Talty, and M. Ames,“Intra-vehicular wireless networks,” in Globecom Workshops, 2007IEEE, 2007, pp. 1–9.

[3] J.-R. Lin, T. Talty, and O. Tonguz, “Feasibility of safety applicationsbased on intra-car wireless sensor networks: A case study,” inVehicular Technology Conference (VTC Fall), 2011 IEEE, 2011, pp. 1–5.

[4] C. Buckl, A. Camek, G. Kainz, C. Simon, L. Mercep, H. Staehle, andA. Knoll, “The software car: Building ict architectures for futureelectric vehicles,” in Electric Vehicle Conference (IEVC), 2012 IEEEInternational, March 2012, pp. 1–8.

[5] M. Eder and A. Knoll, “Design of an experimental platform for anx-by-wire car with four-wheel steering,” in Automation Science andEngineering (CASE), 2010 IEEE Conference on, 2010, pp. 656–661.