Border Crossing Connectivity for CCAM Vehicles: A Field Test at the Norwegian-Swedish Border

Within the EU financed MODI project [1], cross border logistics will be investigated and demonstrated driving from Sweden to Norway at the border of Svinesund (Fig. 1). Border crossing for CCAM (Cooperative, Connected and Automated Mobility) poses several unique challenges. For the envisioned vehicles at SAE Level 4 [2], where a driver within the vehicle is optional, automated driving across the border must be done in a technically and regulatory robust and seamless way. This paper focuses mainly on connectivity and presents results from a field test trial collecting several relevant parameters related to connectivity in multiple runs across the border. Crossing national borders, the LTE/5G communication link will be interrupted because a new roaming mobile network must be selected, and this may create some time without network connectivity. Typically, this happens because the modem tries to use its current network as long as possible. At some point the network quality drops below a threshold, triggering the modem to search for a new network using roaming procedures.

Level 4 vehicles are designed to operate within its Operational Design Domain (ODD). One requirement may be that it must be possible for a remote operator to supervise the vehicle and take control if the vehicle goes outside its ODD at any time. This will require a reliable communication link with adequate bandwidth. Furthermore, tasks related to the actual border crossing may also need connectivity.

Bild vergrößern

Therefore, seamless handover from a network operator to another operator at a border crossing is an important issue to investigate. One solution could be to use alternative communication services for redundancy when crossing the border, including C-ITS communication, relying on the vehicles to have implemented support for hybrid communication. Also, mobile networks are evolving, and new solutions are developed by 3GPP and the telecom industry to utilize 5G capabilities, e.g. [3], where operators cooperate to support seamless handovers between different networks [4]. Cross border roaming and handovers has also been studied in the 5GCroGo project [5]. Fast roaming procedures have been demonstrated at AstaZero in cooperation with Ericsson and Daimler [6], but real condition cross border measurements have not been tested systematically in available published materials.

To test cross border connectivity a test protocol was developed and implemented for Linux and Android to measure relevant quality indicators. The test application sends specially crafted datagram messages to a dedicated server on the internet to facilitate measurement of packet throughput rate and packet loss independently in uplink and downlink directions. Round Trip Time (RTT) was also calculated. On the Android application, parameters such as position, timestamp, signal strength, operator, country code, technology (EDGE, WCDMA, LTE, LTE-CA, 5G-NSA etc.) were measured about every second. Data packet size and rate are configurable independently in uplink and downlink directions. In the tests, uplink packet size 200 with a rate of 2 Hz and downlink packet size 800 bytes with a rate of 4 Hz were used. The analysis focuses on the downlink packet performance as seen from an application using off the shelf equipment.

Several trips with several smart phones each trip was travelled, resulting in 40 registered trips across the Swedish-Norwegian border at Svinesund in both directions. Both high-end Samsung S21 smart phones with 5G connectivity and low-cost Motorola E6 (without 5G) smart phones where used, with SIM cards from all three national operators in Norway (Telia, Ice and Telenor) used as well.

Of the 40 trips registered across the border only one trip showed very little to no connectivity challenges driving across the border. In the following we will look closer into the 39 trips with clear challenges detected. In the analysis we will focus on the three parameters “Download package loss”, “Signal strength” and “RTT”. First of all, we define service loss, or the no service area, as the time period from the first data point with above 90% package loss to the last data point with less than 90% package loss, which for our purpose and application seems appropriate enough. In Fig. 2, the result from one of the trips is shown as an example. In this case we see a total loss of service from about 0.5 km into Norway, lasting for about 3 km or, in this case, 132 s.

Bild vergrößern

In Fig. 2, one can clearly see a significant drop on the signal strength and increase in the RTT as the vehicle is approaching the no service area, and with no valid measurement for these two parameters for some periods inside the no service area. For the measurements we do have inside the no service area, the signal strength is significantly lower and RTT is significantly higher, compared to the outside measurements, suggesting the connectivity is unstable at best. The green crosses in this picture represent changes between mobile network cells. From this pattern we see that the low service quality retrieved inside the no service area (approx. 1 km in) is due to a handover to a new cell, however, even though the vehicle at this point is on Norwegian land, this network cell it connects to is located in Sweden. So, the actual handover between Swedish and Norwegian network are not happening before at the end of the no service area (approx. 3.5 km in) for this case.

Many of the other trips are also showing situations where there is a total loss of service inside the identified no service areas, see Fig. 3 for two examples where the identified no service areas are plotted for two additional trips (one in each direction).

Bild vergrößern

Next, we present some statistics to summarize results from the all measurements, see Table 1 and Table 2 for the distance and time of the identified no service areas, respectively. In the following we will use n to denote the number of observations, µ to denote the mean of some observations (e.g. the distance of the no service area), and ϭ to denote the corresponding standard deviation.

The speed of the vehicle going across the border in the direction of Sweden to Norway is (µ, ϭ) = (17.25, 10) and from Norway - Sweden is (µ, ϭ) = (25.9, 1.2) for all the registered trips. The difference in speed of the two direction is due to congestion being observed for some of the trips going from Sweden to Norway. There seems to be a difference between the distance of the no service area between the two directions (see Table 1), some to be explained by the higher mean speed observed in the direction of Norway to Sweden and some due to topography. Therefore, we pay most attention to the time spent in the detected no service areas in Table 2. Here we also see a difference in the time with no service for the two directions (179 s and 234 s for Swe-Nor and Norway-Sweden, respectively), which indeed is shown to be significant running a two-sided t-test with p-value = 0.0133. Moreover, there cannot be shown a strong correlation between the speed of the vehicle and time in the no service area (p = 0.132 using Pearson's r statistics), suggesting that the difference in distance of the no service area in Table 1 cannot be explained by the congestion and the lower speeds alone. The longest service loss observed was 6 min and 34 s.

Lastly, there is also a significant difference in the standard deviation calculated for both the time and distance. Whereas the direction of Sweden-Norway are showing a lower extent of the no service both in time and distance than for the Norway-Sweden direction, the opposite conclusion is drawn for the standard deviation, where the direction of Sweden-Norway shows a larger variance in both time and distance. There seems to be very little difference between the operators, with the direction of Sweden-Norway for Telenor being an exception with respect to time, however more observations should be gathered to confirm this.

Our results show that there is a significant connectivity service loss when driving across the Swedish-Norwegian border. The identified loss of approximately 200 s would pose a significant challenge for CCAM vehicles depending on connectivity for their operation. There would mainly be three approaches to prepare for such scenarios, one being enabling vehicles with additional sensors and local maps to extend the time with valid ODD without connectivity, and the second being developing and implementing solutions in the mobile network infrastructure to reduce the time with loss of service. For the latter, EU projects are initiated [5]. As a third option, also mentioned in the introduction, would be to use C-ITS in a hybrid solution with LTE/5G at the border.

Moreover, connectivity is not the only challenge when crossing the border with a logistic CCAM vehicle. Other technological challenges with respect to positioning include GNSS correction service and map datum issues. Many countries, including Norway and Sweden, have established national NRTK (Network Real Time Kinematic) services that increasingly are getting attention from the ITS and CCAM community for providing highly accurate corrections for GNSS signals, typically 1–2 cm accuracy under good conditions. To use these highly accurate national services a switch between the national services must be established. However, unique subscriptions for each service are needed and a seamless switch between the two is not an option. The NRTK service is accessed over the internet using mobile networks. As an alternative, several GNSS equipment providers develop systems for cross border corrections, typically through PPP (Precise Point Positioning) or PPP-RTK services, but with slightly less accuracy, typically 3–20 cm under good conditions. The innovation in this area is high, including work from 3GPP, where the telecom industry standardizes broadcasting of correction data via the mobile network [7]. Both commercial companies and international organizations provide real-time PPP correction data, the most recent development is the free service from Galileo High Accuracy Service (HAS).

5G mobile network standards describe network elements and interfaces to support fast handovers and fast roaming between different mobile networks. The work done in [5] shows that the time with service loss can be reduced to a few seconds and even down to order of 100 ms for the ultra-tight coupling between core networks and assistance from the user equipment (modems). Many border crossings are in remote or rural areas where mobile network coverage and capacity is limited. Traditionally, mobile networks near borders have been configured to minimize probability for network reselection to be able to keep network traffic in own network for commercial reasons. Given the complexity of the commercial and technical arrangements needed for seamless roaming it is not always a good business case to invest in seamless cross-border connectivity. Furthermore, regulatory requirements on e.g. frequency plans can make this technically challenging.

In addition, several organizational and regulatory challenges do exist, including traffic regulation, cyber security for traffic rules and regulations, and V2X (vehicle-to-everything) communication, vehicle regulations, and particularly important for logistic operations is custom processes and document handling for CCAM vehicles.