Mode 1 or Mode 2 Sidelink When the Uu Link Goes Gray

The real choice between gNB-scheduled sidelink and autonomous sensing-based allocation is not about peak throughput. It is about what happens after the Uu link starts to die and before the UE admits that it has. Mode 1 gives you collision-free scheduling of large sensor frames and tight retransmission control while the cell is healthy. Mode 2 gives you a sidelink with no upstream dependency, and the decision flips at the point where your network's fallback time becomes longer than the application's survival time.

Where Mode 1 earns its place

In Mode 1 the gNB sees every sidelink buffer status report in the cell, so it can hand out non-overlapping resources and almost eliminate the half-duplex and hidden-terminal collisions that hurt distributed allocation. Periodic object lists fit cleanly onto configured grants. Dynamic grants can carry the large aperiodic bursts of raw or lightly processed sensor data, and that is exactly the traffic Mode 2 sensing handles worst, because a history of reservations tells you nothing about a one-off multi-slot transport block that shows up without warning. HARQ feedback goes back through the gNB, which allocates a retransmission only when one is actually needed, so spectrum is not spent on blind repetitions. In a corridor with engineered, redundant coverage, my estimate is that Mode 1 delivers a few tens of percent more sensor payload under heavy load than a well-tuned Mode 2 pool, and that is a real advantage for high-rate sharing.

The failure that sits between healthy and dead

Mode 1 retransmissions depend on a loop through the cell. The UE reports a sidelink NACK on PUCCH, and the gNB answers with a new DCI grant. When the downlink degrades, grants are missed, and when the uplink degrades, NACKs and buffer reports never arrive. Meanwhile the UE is still counting out-of-sync indications while T310 runs, a timer commonly set to 1000 ms in macro deployments, and only when that timer expires does the UE declare radio link failure and move to the exceptional pool. Through that whole interval the PC5 link between the two vehicles can be perfectly good, yet it starves because the scheduler it depends on can no longer reach it. The exceptional pool then makes things worse in a way that is easy to underestimate. Its random selection without sensing collides with every other UE that lost the same cell, and those losses are correlated, because a tunnel or a failed site drops a whole platoon at the same moment.

The decision rule

To compare the two options, take the worst-case Mode 1 recovery time as the sum of the gray-zone interval, the radio link failure timer and the switch onto the exceptional pool, then set it against the application's survival time, meaning how long the maneuver can run on stale or missing peer data before it has to degrade. If that survival time is shorter than the recovery time on any segment of the operational design domain where coverage is not engineered and redundant, Mode 2 wins at that boundary, and it wins without qualification, because no scheduler tuning can shorten a gap that the protocol itself requires. For cooperative maneuvers my estimate puts survival time at a few hundred milliseconds, which sits below typical RLF detection, so for most highway and urban domains the boundary lands on the Mode 2 side. Mode 2 still pays for this in collisions under bursty, large payloads. Here the release matters: Release 17 inter-UE coordination lets a receiver tell a transmitter which resources it sees as occupied, and that recovers a meaningful part of the hidden-terminal loss without bringing back any dependency on Uu.

What remains open after the choice

Choosing Mode 2 moves the risk into resource pool configuration, because a pool sized for periodic awareness messages will collapse once raw sensor bursts arrive, and that sizing has to be frozen along with the architecture instead of being tuned after the first field trial.

Can we configure both modes on the same UE and get the best of each?

Only if the switch from Mode 1 to Mode 2 does not depend on the Uu link that is failing. If the trigger is an RRC reconfiguration or a declared radio link failure, the dual configuration brings back the same gray zone, so the fallback needs a local trigger, such as missed grants or sidelink HARQ outcomes seen at the UE, before it improves on Mode 2 alone.

Back to NRV2X · Domains

← Back to NRV2X
Engineering & Architecture Brief

Mode 1 or Mode 2 Sidelink When the Uu Link Goes Gray

The real choice between gNB-scheduled sidelink and autonomous sensing-based allocation is not about peak throughput. It is about what happens after the Uu link starts to die and before the UE admits that it has. Mode 1 gives you collision-free scheduling of large sensor frames and tight retransmission control while the cell is healthy. Mode 2 gives you a sidelink with no upstream dependency, and the decision flips at the point where your network's fallback time becomes longer than the application's survival time.

Where Mode 1 earns its place

In Mode 1 the gNB sees every sidelink buffer status report in the cell, so it can hand out non-overlapping resources and almost eliminate the half-duplex and hidden-terminal collisions that hurt distributed allocation. Periodic object lists fit cleanly onto configured grants. Dynamic grants can carry the large aperiodic bursts of raw or lightly processed sensor data, and that is exactly the traffic Mode 2 sensing handles worst, because a history of reservations tells you nothing about a one-off multi-slot transport block that shows up without warning. HARQ feedback goes back through the gNB, which allocates a retransmission only when one is actually needed, so spectrum is not spent on blind repetitions. In a corridor with engineered, redundant coverage, my estimate is that Mode 1 delivers a few tens of percent more sensor payload under heavy load than a well-tuned Mode 2 pool, and that is a real advantage for high-rate sharing.

The failure that sits between healthy and dead

Mode 1 retransmissions depend on a loop through the cell. The UE reports a sidelink NACK on PUCCH, and the gNB answers with a new DCI grant. When the downlink degrades, grants are missed, and when the uplink degrades, NACKs and buffer reports never arrive. Meanwhile the UE is still counting out-of-sync indications while T310 runs, a timer commonly set to 1000 ms in macro deployments, and only when that timer expires does the UE declare radio link failure and move to the exceptional pool. Through that whole interval the PC5 link between the two vehicles can be perfectly good, yet it starves because the scheduler it depends on can no longer reach it. The exceptional pool then makes things worse in a way that is easy to underestimate. Its random selection without sensing collides with every other UE that lost the same cell, and those losses are correlated, because a tunnel or a failed site drops a whole platoon at the same moment.

The decision rule

To compare the two options, take the worst-case Mode 1 recovery time as the sum of the gray-zone interval, the radio link failure timer and the switch onto the exceptional pool, then set it against the application's survival time, meaning how long the maneuver can run on stale or missing peer data before it has to degrade. If that survival time is shorter than the recovery time on any segment of the operational design domain where coverage is not engineered and redundant, Mode 2 wins at that boundary, and it wins without qualification, because no scheduler tuning can shorten a gap that the protocol itself requires. For cooperative maneuvers my estimate puts survival time at a few hundred milliseconds, which sits below typical RLF detection, so for most highway and urban domains the boundary lands on the Mode 2 side. Mode 2 still pays for this in collisions under bursty, large payloads. Here the release matters: Release 17 inter-UE coordination lets a receiver tell a transmitter which resources it sees as occupied, and that recovers a meaningful part of the hidden-terminal loss without bringing back any dependency on Uu.

What remains open after the choice

Choosing Mode 2 moves the risk into resource pool configuration, because a pool sized for periodic awareness messages will collapse once raw sensor bursts arrive, and that sizing has to be frozen along with the architecture instead of being tuned after the first field trial.

Technical FAQ: Can we configure both modes on the same UE and get the best of each?

Only if the switch from Mode 1 to Mode 2 does not depend on the Uu link that is failing. If the trigger is an RRC reconfiguration or a declared radio link failure, the dual configuration brings back the same gray zone, so the fallback needs a local trigger, such as missed grants or sidelink HARQ outcomes seen at the UE, before it improves on Mode 2 alone.

Related Architecture Specifications (AUTOMOTIVEV2X): v2amr.com · v2ecosystem.com · v2road.com

More Technical Articles on nrv2x.com

Deep Dive: NR Sidelink Mode 2 Autonomous Resource Allocation and Collision Resolution →Platooning Maneuver Coordination: Join, Leave, and Emergency Split Handshakes →