The Role of Collective Perception and 5G NR-V2X Sidelink in Road Safety

1

The Role of Collective Perception and 5G NR-V2X Sidelink in Road Safety

arXiv:2609.01478v1 [cs.NI] 1 Sep 2026

Vittorio Todisco, Mattia Andreani, Maria Luisa Merani and Alessandro Bazzi

Abstract—Vehicles and roadside infrastructure are increasingly equipped with sensors capable of perceiving their surroundings. Sharing this information through vehicle-to-everything (V2X) communications is a key enabler of Day-2 applications and is supported by the ETSI collective perception service (CPS). While CPS is expected to play a fundamental role in future intelligent transportation systems, its operation may significantly increase channel load, posing challenges in terms of radio resource utilization, communication reliability, and information management. This paper reviews the current status of CPS standardization and investigates its impact in dense deployment scenarios where connected vehicles communicate through fifth-generation (5G) New Radio-V2X (NR-V2X) sidelink (SL) communications. The main contribution is a realistic evaluation of communication reliability, latency, channel occupancy, and information usefulness under different object-selection strategies for collective perception messages. The analysis is conducted through a networklevel simulation framework integrating empirical object traces derived from real-world datasets, thereby avoiding the limitations of synthetic traffic models. Results show that perception message generation and radio access mechanisms are tightly coupled and should be jointly designed to maximize the benefits of collective perception services. Index Terms—Road Safety, Connected Vehicles, Collective Perception, Vehicular Networks, 5G NR-V2X.

I. I NTRODUCTION

I

N recent years, road infrastructure and vehicles have undergone a radical transformation, driven by the advent of electric propulsion systems and autonomous vehicles. In parallel, an increased attention to sustainable mobility has emerged, spurring the usage of individual means of transportation alternative to passenger cars, such as bicycles and escooters. Still, road fatalities remain one of the worst scourges of our time. Increasing road safety requires a holistic approach that spans from urban redesign for pedestrians and cyclists to the adoption of all technologies that contribute to sheltering the travelers. In this arena, the European Union is among the frontrunners, formally stating the long-term goal to move as close as possible to zero fatalities in road transport by 2050. On the technological front, autonomous or highly automated vehicles and artificial intelligence tools can greatly reduce human errors, contributing to proactive crash prevention in many ways. Yet, a further technology brick is mandatory: vehicles should exhibit cooperation capabilities to overcome the isolated, egocentric view that characterizes even the most Vittorio Todisco and Alessandro Bazzi are with the DEI Department, University of Bologna, Bologna, Italy, and the National Laboratory of Wireless Communications (WiLab), CNIT. Mattia Andreani and Maria Luisa Merani are with the Department of Engineering “Enzo Ferrari”, University of Modena and Reggio Emilia, Modena, Italy, and CNIT.

sophisticated intelligent vehicle. In turn, cooperation relies on connectivity, enabled by vehicle-to-vehicle (V2V), vehicleto-infrastructure (V2I), vehicle-to-network (V2N), vehicle-topedestrian (V2P), and, more generally, vehicle-to-everything (V2X) communications. At present, the deployment of what is known in Europe as the cooperative-intelligent transport system (C-ITS) is still in its infancy. The scenario may change quickly with the advent of collective perception, where vehicles equipped with onboard units (OBUs), as well as road side units (RSUs), broadcast not only their own status but also what they perceive in the surrounding environment. In the literature, existing works on collective perception service (CPS) primarily investigate the impact of perception traffic on radio resource usage by relying on synthetic traces of detected objects and earlier versions of the collective perception message (CPM) generation rules (e.g., ETSI TS 103.324 V2.1.1 or TR 103.562 V2.1.1) [1]–[4]. As a result, two key aspects remain insufficiently understood. First, there is limited knowledge of how many objects are perceived by real vehicles and how this affects the CPM traffic characteristics. Second, perception and communication are treated separately, whereas the CPS effectiveness depends on their interaction, especially in light of the most recent standardization efforts. ETSI specifications (such as those in TS 103.324) are moving toward adaptive message construction mechanisms based on the value of information (VoI), not reflected in prior works. The present study addresses these gaps by integrating object traces from a real autonomous vehicle dataset into a simulation framework, and examining multiple strategies to build perception messages via VoI-based or VoI-agnostic strategies. The novel contributions of this work can be summarized as follows: An updated and contextualized view of CPS standardization, highlighting recent developments and open challenges; • A realistic characterization of CPS traffic by leveraging real-world perception data; • An assessment of the 5G NR-V2X sidelink (SL) performance under different strategies for selecting the objects to embed in the perception messages, including VoIinspired approaches aligned with recent ETSI developments. The evaluation is carried out in terms of reliability, timeliness, channel congestion, and value of the received information. •

The remainder of this article first explains the vehicular communication alternatives and the foreseen evolution of safety services, then discusses collective perception, and finally pro-

2

vides some key performance results when 5G NR-V2X SL is the enabling radio access technology, identifying some open challenges in this research area. II. T HE ROADMAP TO V2X- BASED S AFETY A. Organizations and Standardization Bodies The importance of road safety is reflected by the number of organizations exerting an influence to foster the adoption of vehicular connectivity and build a secure traveling environment in different world regions. Wireless access technologies form the foundation of C-ITS, and the bodies that release standards in this field are the Institute of Electrical and Electronics Engineers (IEEE) and the Third Generation Partnership Project (3GPP). IEEE developed the 802.11p standard for vehicular communications during the first decade of this century, and its enhancement 802.11bd in the latest years; both stemmed from the original 802.11 specifications for wireless local area networks. Since 2016, 3GPP introduced the alternative group of solutions called cellular-V2X (C-V2X), first providing support to vehicular communication via LTE-V2X, and later adding the fifth generation (5G) New Radio (NR)-V2X as a further option. As regards the standards that define the functional aspects, in Europe, the main reference organization is the European Telecommunications Standards Institute (ETSI), whose ITS technical committee develops documents covering the whole communication stack on which the ITS applications rely. In the US, a similar role is played by the Society of Automotive Engineers (SAE), which has the advancement in connected and automated mobility systems among its core functions. Associations also play an important role. Among them, the CAR2CAR Communication Consortium (C2C-CC) and the C-ROADS Platform are key European players supporting the Vision Zero objective through the deployment of C-ITS. C2CCC focuses on V2X communication standards, interoperability specifications, and deployment strategies among automotive stakeholders; C-ROADS primarily includes road authorities and operators involved in harmonizing and testing C-ITS services across borders. A further major actor is the cross-regional industry consortium 5G Automotive Association (5GAA), which has strongly promoted C-V2X-based solutions through white papers and field tests over the past decade. B. Enabling Communication Technologies From the standpoint of access technologies, the first introduced solution is IEEE 802.11p, ratified in 2010 as a variant of the specifications for wireless local area networks, adapted for the highly dynamic vehicular environment; in Europe, it is the basis for ITS-G5. It operates in the 5.9 GHz ITS band with 10 MHz channels, supporting raw data rates ranging from 3 to 27 Mbps. The channel access mechanism is based on a listen-before-talk principle, with unicast and broadcast capabilities. In 2023, IEEE 802.11bd was defined to enhance data throughput, reliability, and latency; it supports wider channel bandwidth (up to 40 MHz), higher-order modulation schemes, and new coding options, with a peak data rate of 200 Mbps

[5]. Crucially, 802.11bd, now part of 802.11-2024 and ITSG5, ensures interoperability with 802.11p, enabling a smooth transition and coexistence in mixed vehicular networks. To date, more than 2 million ITS-G5-equipped vehicles have been sold in Europe. In recent years, 3GPP introduced LTE-V2X in Release 14 (3GPP TS 36.321) and enhanced it in Release 15. It natively supports two communication types: direct SL communications over the PC5 interface (for V2V, V2I, and V2P) and network-based communications over the Uu interface (for V2N). To address the stringent requirements of connected and automated driving, 3GPP later specified 5G NR-V2X in Release 16 (3GPP TS 38.321). 5G NR-V2X redesigns LTEV2X to support periodic and aperiodic traffic, as well as a broader set of communication types (unicast, groupcast, and broadcast). 5G NR-V2X SL employs a CP-OFDM waveform, organizing the time domain in synchronous time slots. In the frequency domain, the available bandwidth is divided into subchannels, consisting of contiguous resource blocks (RBs) which constitute the smallest allocation units. Each transmission is further characterized by a specific modulation and coding scheme (MCS), which determines the combination of modulation order and coding rate used to map bits to physical resources. 5G NR-V2X SL supports two resource allocation modes: in Mode 1, resource scheduling is managed by the cellular base station (gNB), while in Mode 2, vehicles autonomously select resources from a pre-configured pool, enabling the operation outside of cellular coverage. In Mode 2, users select resources after performing a channel sensing phase, then access them according to two alternative mechanisms for distributed resource allocation: sensing-based semipersistent scheduling (SB-SPS) and sensing-based dynamic scheduling (SB-DS). With SB-SPS, a vehicle reserves radio resources at regular intervals for a predefined number of consecutive transmissions. After completing these transmissions, it either reselects new resources with probability 1-P or retains the previous allocation with probability P . If a newly generated message does not fit within the reserved resources, a new selection is triggered. SB-SPS is particularly suitable for constant-size, periodic traffic. In contrast, SB-DS allows vehicles to dynamically select resources based on current traffic conditions, making it more appropriate for irregular or event-driven messages, such as emergency alerts. However, since transmissions in SB-DS are not generally protected by prior reservations, their reliability is typically lower than that of SB-SPS. In China and the US, C-V2X technology is currently being selected as the only solution, with some uncertainty about whether LTE or 5G NR will be adopted. C. The Progression Safety services, built on top of either IEEE or 3GPP radio technologies, are evolving through three main phases, commonly referred to as Day 1, Day 2, and Day 3 of cooperative driving. In Day 1, the paradigm of the connected road user is “I share who and where I am, and how my dynamic attributes are changing”. The broadcasting of messages to stations within radio range augments the vehicles’ perception of the

3

Fig. 1. Example of collective perception for road-safety enhancement. A vehicle (or an RSU) detects a pedestrian and notifies its presence to an approaching car with limited visibility.

environment and builds cooperative awareness. Examples of Day 1 messages include the cooperative awareness messages (CAMs), standardized by ETSI, which convey vehicle mobility updates with a generation rate that adapts to changes in position, speed, and heading, and the decentralized environmental notification messages (DENMs), which carry eventdriven information such as the presence of roadworks or a recently occurred accident. A major challenge for Day 1 safety services is that their efficacy depends strongly on the penetration rate of connected vehicles and infrastructures. Moreover, although specific messages broadcast by vulnerable road users (VRUs), called VRU awareness messages (VAMs), have been defined, most pedestrians and cyclists are expected to remain unconnected, revealing the intrinsic limitation of cooperative awareness. In Day 2, the paradigm shifts to “I share what I see”, that is, to collective perception [1], [2]. In this phase, connected vehicles equipped with sensors such as cameras, lidars, and radars, broadcast information about their locally perceived environment. This information-sharing approach represents a powerful and transformative step toward enhanced road safety for several reasons: (i) Connected road users receiving perception messages can fuse the shared data with their own sensor measurements, thereby extending their perception range beyond the local sensor view; (ii) Unconnected VRUs can be detected and their presence signaled to connected vehicles; (iii) The effectiveness of collective perception is less dependent on the penetration rate of equipped vehicles compared to cooperative awareness, since vehicles or RSUs share what they perceive in the surroundings, including non-connected entities, enabling a broader situational awareness. An example scenario where the Day 2 approach can significantly improve safety is illustrated in Fig. 1. Collective perception is currently one of the most prominent research and standardization themes in the road safety domain. In Europe, the format and generation rules of CPMs have been discussed for several years within ETSI; however, the recently released standard is undergoing a substantial revision, as evidenced in the next section. Similar efforts are emerging globally: several pilot projects are underway in China; Korean

technical reports refer to “Extended Sensor Sharing Messages” as the local equivalent of CPMs; and comparable initiatives are being developed in the United States. Once Day 2 applications reach large-scale deployment, the next step will be Day 3, characterized by the paradigm “I share what I plan to do and cooperate for safe and efficient maneuvers”. This phase will be enabled through the exchange of messages called maneuver coordination messages (MCMs) in Europe. As detailed in ETSI TR 103 578 V2.1.1, MCMs will be used for agreement-seeking or prescribed operations: in the former case, vehicles negotiate their intended trajectories to reach a cooperative agreement, whereas in the latter they execute maneuvers prescribed by the infrastructure or another vehicle. A wide range of use cases can benefit from cooperative maneuvering, including lane merging and changing, intersection crossing, and platooning. III. C OLLECTIVE P ERCEPTION A. The C-ITS Stack and the Collective Perception Service As the TCP/IP suite is unfit for safety-of-life applications, the dedicated C-ITS protocol stack was defined by ETSI in EN 302 665. The stack includes three layers: from bottom to top, the access, networking & transport, and facilities layer. The access layer is responsible for transmitting packets over the wireless medium, employing either the IEEE or 3GPP technologies. On top of it, the networking & transport layer includes the GeoNetworking (GNW) protocol, which manages the addressing of nodes and the forwarding of packets in the case of multi-hop communications, and the Basic Transport Protocol (BTP), which is in charge of multiplexing the different flows received from the layer above. The last layer is the facilities layer, whose main components are the services. In C-ITS, services are autonomous or application-controlled software entities that exchange data via structured messages. Unlike typical systems where applications handle data directly, C-ITS services mainly generate and use data themselves and store it in dedicated databases, such as the local dynamic map. The main Day 1 services are the cooperative basic awareness service, which continuously generates CAMs, and the decentralized environmental notification service, which produces event-triggered DENMs. In the case of Day 2 collective perception, the facilities layer includes the CPS, which continuously generates CPMs. The rules for message generation and object inclusion should balance a complex trade-off to ensure effectiveness while avoiding channel overload, as further discussed below. B. Collective Perception Messages ETSI mandates that the information shared by the CPS is conveyed in the form of CPMs. The CPM format includes a header and a payload made of several containers, as illustrated in Fig. 2. The management container provides the reference time and position of the transmitting station, along with segmentation data used when messages are fragmented. The CPM container is a sequence of wrapped containers, each carrying different information. The first wrapped container describes the originating station. For a vehicle, it reports the

4

CPM header

Mandatory container Optional container

payload

Alternatives

6 bytes

managementContainer

cpmContainer

21–24 bytes one among these four

wrappedCpmContainer

...

up to 8 wrappedCpmContainer PerceivedObjectContainer

OriginatingVehicleContainer 3–13 bytes

PerceptionRegionContainer

or

OriginatingRSUContainer 1–2 bytes

1 byte + 8–22 bytes per region

SensorInformationContainer 1 byte + 2–16 bytes per sensor

numberOfPerceivedObjects max 255 1 byte

PerceivedObject

perceivedObjects

...

PerceivedObject

13–66 bytes per object

Fig. 2. Structure of the CPM defined by ETSI TS 103 324 V2.1.1 [6], including the header, payload, and mandatory and optional containers.

bounding-box dimensions, driving direction, speed, acceleration, and heading. For an RSU, it provides the identifier of the intersection or road segment where it is located. Another wrapped container may include sensor information detailing the station’s onboard sensing equipment: the number and types of sensors, identifiers, and detection areas. In a further wrapped container, an optional perception-region container specifies the deviations from the nominal sensing capabilities. One final wrapped container lists each perceived object with its identifier, type, and position, along with optional attributes such as the detecting sensors, confidence levels, age, and kinematic properties (e.g., speed, acceleration). This container is central to our analysis, as it largely determines the resulting CPM size. C. The Issue of Redundancy Mitigation The transmission periodicity of CPMs ranges from 100 ms to 1 s. According to ETSI 103 324 CPM specification, up to version 2.1.1, a message is transmitted when the prescribed object-inclusion conditions are met. VRUs are reported based on the time elapsed since the last inclusion in a message, whereas for vehicles and motorcyclists, the reporting also depends on their dynamic behavior, mirroring the logic used in CAMs. However, these choices may result in excessively large messages and a high channel load. For instance, vehicles in a congested area may perceive nearly the same large number of objects and redundantly report all of them in the messages. This issue motivated the study of redundancy mitigation techniques [3], [4], which omit objects included too frequently, with limited motion, poor local confidence level, or too far from the disseminating station. Recently, the redundancy problem has prompted a substantial reconsideration of the message generation rules within ETSI. The early 2026 draft revision of ETSI TS 103 324 gives VoI a prominent role in CPM construction, indicating that objects with higher VoI should be prioritized for inclusion in the message, while the number of advertised objects depends on the communication resources available for transmission.

The VoI of an object may depend on several factors, such as its proximity to the road user, the unpredictability of its motion, or the timeliness and reliability of its information. For instance, a pedestrian suddenly stepping onto the road, as in Fig. 1, has a much higher VoI than a stationary parked car. By focusing on the most valuable data, vehicles and RSUs share fewer but more meaningful messages, reducing the load on the communication channel while improving collective situational awareness. However, different driving environments call for different interpretations of what the term “valuable” means. Computing the VoI of an object is a challenging research question, since it is application-dependent, and its definition varies in different contexts [7]. For vehicular networks, some proposals have recently emerged in literature [8]–[10]. IV. R ESULTS FROM R EAL DATA In this section, we present the findings of our investigation into the capability of 5G NR-V2X to deliver collective perception traffic. To ensure a realistic evaluation, we employ object traces derived from real sensor measurements collected by moving vehicles, and conduct simulations using the opensource WiLabV2Xsim framework [11]. The analysis focuses on a congested highway scenario in which a fleet of CPScapable vehicles share information about a large number of surrounding objects employing Mode 2 SB-SPS. A. Perceived Objects from Real Sensor Datasets A realistic characterization of collective perception traffic requires the use of actual sensor data. To this end, we employed the Cirrus dataset [12], an open-source collection of annotated scenes sampled at 1 Hz by a vehicle driving through highways and urban streets located in Palo Alto, California. The recording vehicle was equipped with an RGB camera, two lidars, two GPS units, and one IMU. An illustrative frame captured by the vehicle camera is shown in Fig. 3(a), referring to a congested highway scenario. Although Cirrus and similar datasets are typically used in computer vision tasks such as pattern and image recognition, here we leverage Cirrus

5

B. Object Inclusion Strategies We examine three strategies for object inclusion: (i) Baseline: all the detected objects are included in the CPM; (ii) Random inclusion: only a subset of the perceived objects is included in the message, with each object independently selected according to a probability pincl , pincl < 1; (iii) VoI-rank inclusion: only the perceived objects with the highest VoI value are included, subject to the constraint that the message does not exceed a predefined size, Smax . The first strategy serves as the reference, while the second captures how reducing the CPM size via the pincl value affects system performance. The third solution follows the new approach in the ETSI CPS standard, where the objects are ranked based on their VoI and then those with the largest VoI are selected until a maximum message size is reached; this maximum size is assumed fixed, but may be dynamically provided, for example, by a radio resource management entity such as the one that is under development in ETSI [14]. Following [15], given an object is at distance d from the detecting vehicle, the VoI of the object is computed according to a linear distance-based law with a negative slope as long as d ≤ dmax , whereas it is 0 if d > dmax , with dmax = 50 m.

Communication performance is assessed in terms of: (i) packet reception ratio (PRR), defined as the ratio between

(a) Example camera frame recorded in the Cirrus dataset [12].

Number of Objects

60

Baseline Random incl. VoI-rank

40 20 0

0

20

40

60

80

100

Time [s]

(b) Objects included in a CPM as a function of time. Tgen = 200 ms.

Packet Size [B]

content to address a different question: What is the number of perceived objects and their variability over time? This will serve as an input to characterize the perception traffic. In the dataset, a recorded scene consists of a sequence of annotated frames, and every annotation corresponds to an object represented by a bounding box. By tracking the evolution of the bounding boxes over time, the objects entering and leaving the vehicle’s field of view are identified. These correspond to the objects the vehicle detects [13], whose data may or may not be included in the CPM depending on the examined message-generation strategy. Given the 1 Hz sampling rate of the dataset, the frames provide the number of objects a vehicle would include when CPMs are generated every second. To investigate shorter generation periods (e.g., 200 and 400 ms), a linear interpolation of the number of detected objects and their distance over time is applied. To determine the size of the packet transmitted over the radio channel, the relevant CPM containers shown in Fig. 2 are considered, then the ITS headers are added, as discussed in subsection III-A. Namely, the CPM includes a 6-byte header, a 21-byte managementContainer, a 5-byte OriginatingVehicleContainer, and a 10-byte SensorInformationContainer (updated every second). The PerceptionRegionContainer is omitted. The PerceivedObjectContainer contributes 13 bytes per detected object, providing basic information. The BTP and GNW headers attached to the CPM are 4 and 12 bytes long. A 231-byte full security certificate is further included in the packet every 1 second; when the CPM periodicity is lower than 1 s, the 89-byte digest is included in all remaining packets.

# Subch. >5 Baseline Random incl. VoI-rank 5

1000

4 3

500

2 0

1 0

20

40

60

80

100

Time [s]

(c) Packet size and required subchannels over time. Tgen = 200 ms. Fig. 3. Starting from the object annotations in the dataset, the figure shows how different object-inclusion strategies affect the CPMs generated over time by the sensing vehicle. The number of included objects, packet sizes, and required subchannels are derived from the real perception trace.

the number of receiving vehicles successfully decoding a packet at a given distance from the transmitting vehicle and the total number of vehicles located at that distance; (ii) packet inter reception (PIR), defined as the time between two consecutive successful receptions of packets belonging to the same application flow (as per ETSI TR 101 613); (iii) channel busy ratio (CBR), defined as the fraction of subchannels with a Received Signal Strength Indicator (RSSI) above -88 dBm (ETSI EN 303 798 - V2.1.1), within a 100 ms interval; this threshold is approximately 20 dB above the ideal thermal noise floor and indicates the presence of strong interference; we compute the median of the values obtained from all stations in all simulation intervals. To estimate perception effectiveness, we introduce the received VoI ratio (RVR). Following ETSI definition, the transmitted VoI of a CPM is obtained as the sum of the VoI values of the objects included in the message. The corresponding received VoI at distance D is estimated as the transmitted VoI multiplied by the average PRR at that distance. The RVR is then obtained by averaging, over all CPMs, the received VoI normalized against the full-information case where all perceived objects are included every 100 ms and always correctly received. Hence, the RVR represents the average fraction of the full-information reference that is delivered at a given distance. It captures, in a single metric, object selection, CPM generation frequency, and communication reliability.

6

Baseline, Tgen = 100 ms Baseline, Tgen = 200 ms Baseline, Tgen = 400 ms Random incl., Tgen = 200 ms VoI-rank, Tgen = 200 ms 1

Pr{PIR ≥ x}

PRR

0.95

10−1

0.8

0.9

0.85

200

400

RVR

1

1

10−2

0.6 0.4

−3

10

10−4

0.2 0

0.5

1

1.5

2

0

200

400

D [m]

x [s]

D [m]

(a) Packet reception ratio

(b) PIR complementary CDF

(c) Received VoI ratio

Fig. 4. Performance of the object-inclusion strategies: PRR, CCDF of the packet inter-reception time, and received VoI ratio.

C. Simulation Settings and Results In the simulated scenario, vehicles travel along a highway and are randomly distributed across six lanes, three in each direction. The average vehicular density is 150 vehicles/km. The speed of each vehicle is drawn from a Gaussian distribution with mean 70 km/h and standard deviation 7 km/h; the mean corresponds to the estimated value in the congested setting recorded in the dataset. Moreover, all vehicles are assumed to be capable of transmitting and receiving CPMs. The SL channel is exclusively dedicated to CPS traffic, in alignment with envisioned real-world deployments. At physical layer, the transmission power is set to 23 dBm, so that the power density depends on the number of subchannels the packet transmission requires. The antenna gain is 3 dBi and the noise figure is 9 dB. Following the ETSI indications, the SL channel is 20 MHz wide and the subcarrier spacing (SCS) is 30 kHz, with at most 5 subchannels available per slot to transmit a single packet (see Fig. 3(c), right axis). The MCS index is set to 12, which corresponds to 16-QAM with coding rate 0.43, and allows a maximum packet size of 1180 bytes without fragmentation. The propagation channel follows the ECC Report 68 rural model, recently adopted for highway scenarios in ETSI TR 103 439. We examine SB-SPS when the keep probability is set to P = 0.8, and no retransmissions are considered. To handle the packet size variability, the allocation of subchannels matches the initial packet size. If the next packet requires fewer subchannels, padding is applied; if it requires more subchannels, a reallocation is performed. The temporal evolution of the number of objects and packet size is provided in Fig. 3. In particular, Fig. 3(b) showcases N , the number of objects included in each message as a function of time, for the three strategies: the baseline, the random inclusion when pincl = 0.5, and the VoI-rank inclusion, when the latter employs a maximum packet size Smax = 400 bytes, corresponding to the largest size that fits in two subchannels. The periodicity at which CPMs are generated is Tgen = 200 ms. Notice that for the VoI-rank inclusion strategy, N toggles between two values: this is due to the transmission of the full security certificate every second, i.e., every five packets.

Baseline, Tgen = 100 ms Baseline, Tgen = 200 ms Baseline, Tgen = 400 ms Random incl., Tgen = 200 ms VoI-rank, Tgen = 200 ms 0.8 600 0.6 400

0.4

200

0.2 0

0 (a) Median CBR.

(b) Shared object perceptions per sec. [s−1 ]

Fig. 5. Channel busy ratio and shared objects per second.

which intermittently limits the N value. This phenomenon is not observed in the two other strategies, as they are not subject to the constant packet size constraint. Fig. 3(c) reports the packet size evolution over time: for the VoI-rank inclusion scheme, it is equal to the maximum or differs from it for very few bytes. This is due to the congested highway conditions considered. On the right-hand side of this figure, the number of subchannels requested for packet transmission is shown, evidencing that it progressively reduces as stricter inclusion rules are introduced. Since the dataset contains objects from a single recording vehicle, each vehicle in the simulation is assigned a random starting point along the object trace. Once the end of the trace is reached, the sequence is read in reverse order, and this alternating forward–backward process continues until the end of the simulation. The impact of communication is addressed in Fig. 4. Fig. 4(a) reports the PRR as a function of the transmitter– receiver distance for the considered object-inclusion strategies. The baseline with Tgen = 100 ms disseminates all perceived objects most frequently, but exhibits the fastest PRR degradation with distance. Increasing Tgen to 200 or 400 ms improves the PRR range at the price of less frequent perception updates, while the VoI-rank strategy with Tgen = 200 ms further benefits from a limited and stable packet size. Fig. 4(b) complements this view through the PIR complementary cumulative distribution function (CCDF) for distances below 300 m. The results show the trade-off between frequent updates and reception gaps, with the VoI-rank strategy

7

limiting the probability of large PIR values. Fig. 4(c) reports the RVR, which combines reliability with the value of the delivered perception content. The 100 ms baseline achieves the highest RVR close to the transmitter, where the PRR is high and all perceived objects are included at the reference generation period (100 ms), but its value decreases rapidly with distance. The 200 ms and 400 ms baselines are penalized by their lower generation frequency, while random inclusion is additionally affected by the reduced transmitted VoI. The VoI-rank strategy provides a more stable RVR over distance, since it limits the message size while prioritizing objects with larger VoI. Finally, Fig. 5 provides additional information in terms of channel occupancy and shared objects. The median CBR in Fig. 5(a) confirms that less frequent or smaller CPMs reduce channel congestion. Conversely, Fig. 5(b) shows that reducing the channel load generally also reduces the number of objects shared per second, confirming that the object selection should be guided by a relevance criterion. Overall, the results show that the most suitable strategy depends on the balance among selected information value, message generation frequency, communication reliability, and channel occupancy. V. C ONCLUSION AND F UTURE D IRECTIONS This work reviewed the status of CPS standardization, investigated the impact of CPM traffic on communication performance, and evaluated the resulting perception effectiveness, considering 5G NR-V2X SL as the underlying communication technology. Real-world object traces were integrated into a network-level simulator to first assess communication reliability, timeliness, and resource occupancy, when different strategies for including object information in the CPMs are employed. The conventional communication metrics were then paired by a simple estimate of the amount of valuable information that CPMs convey to receiving vehicles. The results indicate that variations in message size significantly affect the performance of 5G NR-V2X SL and that the VoI-rank inclusion strategy yields substantial communication gains by shaping perception traffic in line with current ETSI standardization. The latter scheme also exhibits a remarkable delivery of valuable information, with limited loss increasing the distance. In the proposed approach, the access layer notifies resource availability in terms of packet size, thereby constraining the message dimension at the facility layer, while the CPS independently ranks objects based on their relevance and fills messages accordingly, thus simplifying system design. Several avenues remain for future research. First, the optimal message size should be determined for different configurations of the physical and access layer, and under different CPM traffic conditions. As a second point, the definition of the value of information assigned to the perceived objects requires additional effort to accurately capture a broader range of aspects, including accuracy, timeliness, and relevance to the receivers. Moreover, future research necessitates novel datasets to quantify performance in alternative settings, such as urban

areas. Finally, large-scale field experiments are still required before practical deployment can be pursued. ACKNOWLEDGMENTS This work was supported by the Italian National projects MoVeOver and ALERT of the “RESTART” program (PE00000001) and by the DIEF FAR 2025 research grant. The authors sincerely thank the reviewers for their insightful comments, which greatly enhanced the quality of this work. R EFERENCES [1] H.-J. Gunther, O. Trauer, and L. Wolf, “The potential of collective perception in vehicular ad-hoc networks,” in ITST, 2015. [2] H.-J. Günther et al., “Realizing collective perception in a vehicle,” in IEEE VNC, 2016. [3] G. Thandavarayan, M. Sepulcre, and J. Gozalvez, “Redundancy mitigation in cooperative perception for connected and automated vehicles,” in IEEE VTC2020-Spring, 2020. [4] Q. Delooz et al., “Analysis and evaluation of information redundancy mitigation for V2X collective perception,” IEEE Access, vol. 10, pp. 47 076–47 093, 2022. [5] G. Naik, B. Choudhury, and J.-M. Park, “IEEE 802.11bd & 5G NR V2X: Evolution of radio access technologies for V2X communications,” IEEE Access, vol. 7, pp. 70 169–70 184, 2019. [6] ETSI, “Intelligent Transport System (ITS); Vehicular Communications; Basic Set of Applications; Collective Perception Service; Release 2,” Tech. Spec. 103 324 V2.1.1, Jun 2023. [7] F. Alawad and F. A. Kraemer, “Value of information in wireless sensor network applications and the IoT: A review,” IEEE Sensors Journal, vol. 22, no. 10, pp. 9228–9245, 2022. [8] M. Giordani et al., “A framework to assess value of information in future vehicular networks,” in Proceedings of the 1st ACM MobiHoc Workshop on Technologies, MOdels, and Protocols for Cooperative Connected Cars, 2019. [9] T. Lyu et al., “Accuracy and relevance: A value of information based prioritisation of perceived objects for the etsi collective perception service,” in IEEE VNC, 2025. [10] V. A. Wolff et al., “Uncertainty and prioritization: Empirical evaluation of a VoI-based CPM generation pipeline using real-world data,” in IEEE VNC, 2025. [11] V. Todisco et al., “Performance analysis of sidelink 5G-V2X mode 2 through an open-source simulator,” IEEE Access, vol. 9, pp. 145 648– 145 661, 2021. [12] Z. Wang et al., “Cirrus: A long-range bi-pattern LiDAR dataset,” in IEEE ICRA, 2021. [13] M. Andreani, L. Lusvarghi, and M. L. Merani, “A statistical characterization of the actual cooperative perception messages and a generative model to reproduce them,” in IEEE FNWF, 2023. [14] A. Bazzi et al., “Multi-channel operation for the release 2 of ETSI cooperative intelligent transport systems,” IEEE Communications Standards Magazine, vol. 8, no. 1, pp. 28–35, 2024. [15] M. Sepulcre, J. Tortosa-Garcia, and J. Gozalvez, “Demand- and priorityaware adaptive congestion control for heterogeneous V2X service requirements,” in IEEE VTC2026-Spring, 2026. Vittorio Todisco [M] (vittorio.todisco@unibo.it) is a research fellow at the University of Bologna, where he obtained his Ph.D. His work focuses on connected vehicles and standardization activities in the ITS sector. Mattia Andreani [GSM] (mattia.andreani@unimore.it) is a Ph.D. student in ICT at the University of Modena and Reggio Emilia. His research interests include vehicular communications and vehicular collective perception. Maria Luisa Merani [SM] (marialuisa.merani@unimore.it) is an Associate Professor at the University of Modena and Reggio Emilia. Her research interests focus on vehicular networking and road safety. She served as an Editor of the IEEE Transactions on Wireless Communications and as General Chair of the 2025 IEEE WCNC conference. Alessandro Bazzi [SM] (alessandro.bazzi@unibo.it) is an Associate Professor at the University of Bologna. His research interests are mainly on medium access control and radio resource management of networks of connected and autonomous vehicles. He is/had been part of the ETSI STFs on C-ITS.

Related documents

• Comprehensive analysis method of key risk factors in the subway operation accident by complex network and accident data#1001740
• Exact Payload-Decoupling Conditions for Pilot-Only BEM Channel Estimation With Application to OTFS#820934
• Safety Blind Spot in Remote Driving: Considerations for Risk Assessment of Connection Loss Fallback Strategies#987413
• Intelligent Sensing for Transportation Safety: Technologies, Innovations, and Challenges.#415554
• Transport safety indicators#139975