Sep 9, 2020
Qualcomm products mentioned within this post are offered by Qualcomm Technologies, Inc. and/or its subsidiaries.
In previous blog posts — and as rigorously shown in laboratory and field tests conducted by Ford and Qualcomm Technologies which were vetted by 5GAA (See 5GAA V2X Functional and Performance Test Report) — we have described the C-V2X link performance advantage and the reasons behind this. In recent months, we have extended our C-V2X testing as part of the Crash Avoidance Metrics Partners (CAMP), LLC consortium. Specifically, CAMP conducted a battery of tests to show C-V2X performance in the field with sets of vehicles driving hundreds of miles on real roads to examine how C-V2X performs under highly congested conditions. (See C-V2X Performance Assessment Project for the CAMP report.)
Why this field work with CAMP? It is the emphatic final step in helping prove that the radio access technology we call C-V2X can deliver vehicle-to-vehicle and vehicle-to-infrastructure safety communications performance in the driving conditions you foresee. Real roads with cars outfitted with C-V2X short range radios is what you’d expect as well as safety performance in difficult environments. In short, C-V2X should perform the important function of communicating safety information to nearby cars when hundreds of vehicles within a few hundred meters are broadcasting C-V2X at the same time. Yes, this is performance under congestion, but not just traffic congestion — although there is a strong relationship. From a radio communication perspective, we’re talking about performance under channel congestion, when many C-V2X radios are simultaneously competing for “air time.” This video (link TBD) illustrates our point.
One way to get good performance under congestion is to decrease the rate that messages are sent, but not so much that it negatively affects fundamental safety communication performance. This allows the standard Basic Safety Message to get to nearby vehicles on time. The protocol on how to do this was developed for different technologies, and in the U.S., it is based on two factors working in opposite: (1) what we call target tracking error, or in simple terms, if the other vehicle is showing a trajectory that’s diverging too much from its broadcast predicted short term path, that error needs to be addressed by giving your car’s message higher priority, and (2) the density of cars in the nearest 100 meters, as high density in the local surroundings means that your car should maybe back off a bit. In short, target tracking error means higher priority, modulating the local density. This protocol has existed for many years, and those of us who are pressing to see C-V2X on roads decided that we should adopt it. Others in the ITS community agreed with us.
Naturally, we wanted to be prepared for this CAMP work, so we decided to conduct our own work in advance of this testing with vehicle manufacturers. We conducted those tests and reported on them in our own C-V2X Congestion Control Study, and we provide test highlights below.
How we tested C-V2X in congested conditions
To demonstrate C-V2X congestion control, scalable test setup architectures both in the lab and in the field were designed to allow flexibility in testing. We used what we call the Vehicular Congestion Test Rack (VeCTR) in our lab to enable focused study of one host and one remote vehicle that experience a concurrent channel load generated by up to 576 emulated background devices. Then we went to the parking lot
s of the SDCCU Stadium (which you might remember as Qualcomm Stadium, the former home of the San Diego Chargers) and tested an elaborate setup that deployed eight moving vehicles, and 50 stationary background devices with each one broadcasting in what we call a “super UE” mode, thereby emulating up to 250 background devices on a long test track.
To assess key performance indicators (KPIs) such as the packet error rate (PER), inter-packet gap (IPG) and inter-transmission time (ITT) were measured to provide a comprehensive view of system performance. These KPIs describe the “goodness” of a sequence of packets for vehicular safety applications and constitute the accepted radio access technology performance parameters necessary to deliver these short-range broadcast messages. As indicated earlier, they are based on extensive legacy research by the Intelligent Transportation Systems (ITS) community. This research is standardized by SAE.
Note that information age (IA) is not used. Many scenarios showed that congestion control might push IA up to 600ms while tracking error remains acceptable for cars broadcasting safety messages. While IA is more easily measured, it should not be the main KPI for evaluating the performance of a safety messaging protocol for vehicle-to-vehicle (V2V), since what really matters are vehicles that encounter high tracking error
Our research showed that the SAE congestion control was triggered and effective in lowering PER and IPG, while ensuring vehicle safety communication between C-V2X nodes in congested scenarios. This is also validated by results that were compared to simulation results for additional validation.
In particular, our VeCTR lab results show the effectiveness of using SAE congestion control with C-V2X to reduce the system load, and enable satisfactory communication at a median PER of 3 percent between a home and remote vehicle at 75m, which corresponds to an approximately 95 dB path loss from each other.
The tests were extended later to show good performance with up to an additional 15 dB path loss between the two devices, and differentiated performance for critical BSMs versus normal BSMs. This initial set of results also validates the simulation results presented to 5GAA for a more extensive 1,940 car scenario (See 5GAA V2X Functional and Performance Test Report). The simulations and lab results match at the PER and CBR levels. As described in the 5GAA report, the main contributors to channel congestion comes from the vehicles most proximal to the host vehicle. We observed that with 250 vehicles, the congestion control protocols triggered are sufficient (to counter the argument that we need more vehicles to realistically assess the C-V2X operation in congested environments).
The field test setup that generated results for this report found that 50 background devices generated similar loads to the VeCTR setup on 300m or 600m test tracks. A variety of scenarios were tested, including 4 host vehicles and 4 remote vehicles, either mobile or stationary, at various points along the track.
In summary, the tests we conducted helped validate that:
- C-V2X works reliably in congested environments in laboratory and field tests.
- C-V2X is ready for commercial deployments by testing Day 1 safety use cases in simulated real-world situations.
- SAE congestion control protocols work well with C-V2X.
We therefore felt confident that we were ready for the CAMP work referenced at the start of this article. With the CAMP tests under our belt, we are confident in C-V2X as a system, as we have addressed very well how C-V2X performs — and in essence behaves — under severe congestion up to the exact same engineering scrutiny that other technologies have been studied. As a final note, we point out that subsequent independent work performed by CAMP came to the same technical conclusion as described here. C-V2X is designed to to reliably ensure basic safety communication in all environments, including congested real-world scenarios involving hundreds of vehicles.