veins_ros_v2v_ucla

V2V Sensor Data Sharing (SDSM) dissemination benchmark on a UCLA-area SUMO network, using Veins (OMNeT++ + SUMO) with optional ROS 2 bridge. The stack uses full IEEE 802.11p PHY/MAC (path loss, fading, interference, CSMA/CA) instead of toy drop/delay models.

Canonical paper comparison (v2): Periodic, Greedy_v2, and HybridSDSM_v2 at ~400 vehicles, 300 s, seed 0 (effective num_vehicles in *-metadata.csv may be slightly below the requested -n due to SUMO insertion dynamics). See CLAUDE.md for the latest headline numbers and REWRITE_SUMMARY.md for design provenance.

Simulation stack

Component	Role
SUMO	Traffic: car-following, lanes, signals on the UCLA-area net.
OMNeT++	Discrete-event core, modules, `.ini` / `.ned`.
Veins	TraCI bridge, `Mac1609_4`, `Decider80211p`, mobility.
ROS 2 (optional)	UDP bridge for live TX/RX; not needed for batch CSV runs.

Each TraCI step (0.1 s) SUMO advances vehicles; Veins syncs OMNeT++ Car modules. RosSDSMApp decides when to send and which objects to pack, then the MAC/PHY delivers or drops packets. Successful receptions are logged to CSV under results/ (see .gitignore: raw simulations/results/ is local scratch).

Simulation world (current defaults)

Aspect	Setting
Playground	1300 m × 1000 m (`simulations/omnetpp.ini`)
Channel	`config.xml`: `SimplePathlossModel` α = 2.75, Nakagami m = 1.5 (constM); 5.89 GHz decider
TX power	100 mW (20 dBm)
Noise / sensitivity	−98 dBm noise floor, −110 dBm min power (typical Veins 802.11p)
Interference range	`maxInterfDist = 1500 m` (avoids clipping far interferers)
Obstacle shadowing	Off (no building polygons); metadata records `obstacle_shadowing,false`
Fleet size	Set via `run_experiments.py --num-vehicles N`; metadata `num_vehicles` is the effective count for that run

Implementation parameters (reference)

Authoritative definitions: simulations/apps/RosSDSMApp.ned (all appl.* defaults), simulations/omnetpp.ini (world + PHY + per-[Config] overrides), simulations/config.xml (analogue models), sumo/scenario.sumo.cfg (SUMO horizon / step).

Experiment runner (`run_experiments.py`)

Symbol / flag	Typical value	Notes
`SIM_DURATION_S`	300	Default wall-clock sim horizon unless overridden
`DEFAULT_NUM_VEHICLES`	400	Canonical density; routes regenerated via `sumo/gen_ucla_routes.py` → `sumo/routes.rou.xml`
`--seed` / `runnumber`	e.g. 0	Drives `.node[].appl.runNumber = ${runnumber}` and reproducibility
`SEEDS`	`0..9`	Full multi-seed sweep when you run without narrowing `--seed`

SUMO (`sumo/scenario.sumo.cfg`)

Parameter	Value
`begin`	0
`end`	300 (s)
`step-length`	0.1 (s)

OMNeT++ network & TraCI (`simulations/omnetpp.ini` `[General]`)

Parameter	Value
`network`	`networks.TwoCarsScenario`
`sim-time-limit`	300 s
`seed-set`	`${runnumber}`
`repeat`	1
`*.manager.updateInterval`	0.1 s
`*.world.playgroundSize{X,Y,Z}`	1300 m, 1000 m, 50 m
`*.connectionManager.maxInterfDist`	1500 m
`.node[].nic.phy80211p.noiseFloor`	−98 dBm
`.node[].nic.phy80211p.minPowerLevel`	−110 dBm
`.node[].nic.mac1609_4.txPower`	100 mW

PHY analogue models (`simulations/config.xml`)

Model	Parameters
`SimplePathlossModel`	α = 2.75, thresholding on
`NakagamiFading`	m = 1.5, `constM = true`
`Decider80211p`	5.89 GHz center frequency

`RosSDSMApp` — mode switches (NED defaults; override in `omnetpp.ini`)

Parameter	Default	Role
`sendInterval`	1 s	Base interval; [Config Periodic] sets 0.1 s
`periodicEnabled`	true	Periodic path vs greedy-driven
`greedyEnabled`	false	v1 weighted-sum greedy
`hybridEnabled`	false	v1 / v2 hybrid
`hybridVariant`	`"v1"`	`"v2"` for HybridSDSM_v2
`greedyVariant`	`"v1"`	`"v2"` for Greedy_v2 (shared scheduler with Hybrid v2)
`bsmImpliedMode`	false	Header-only / zero-object ablation
`useVoiObjectSelection`	false	Legacy VoI object ranking (v1 hybrid path)
`rosBridgeMode`	`"off"`	`"off"` / `"log"` / `"live"`

Greedy / hybrid v1 utility (NED + `[General]` in `omnetpp.ini`)

Parameter	Default (NED)	Typical `[General]`
`greedyTickInterval`	0.1 s	0.1 s
`greedyAlphaPos` / `greedyAlphaSpeed` / `greedyAlphaHeading`	1.0 / 0.5 / 0.1	same
`greedyW1` / `greedyW2` / `greedyW3` / `greedyW4`	1.0 / 0.5 / 0.3 / 0.0	W4 often 0.5 in `[Config Greedy]`
`greedyThreshold`	1.0	same
`greedyMinInterval` / `greedyMaxInterval`	0.2 s / 5.0 s	same
`congestionWindow`	1.0 s	same
`cbrEwmaAlpha`	0.3	(NED only; not duplicated in ini)
`redundancyEpsilon`	0.5	receiver-side sender-state redundancy
`hybridThreshold`	1.2	v1 hybrid
`hybridRedundancyWindow`	1 s	v1 sender object resend suppression
`hybridWSelf` / `hybridWTime` / `hybridWCBR` / `hybridWObj`	1.0 / 0.4 / 0.4 / 0.6	v1 hybrid utility weights
`hybridVoiDistWeight` / `hybridVoiAgeWeight` / `hybridVoiRelSpeedWeight`	0.6 / 0.3 / 0.1	v1 VoI terms
`hybridMinVoi`	0.0	v1 VoI floor

Multi-object SDSM caps & sensing window

Parameter	`[General]`	`[Config Greedy_v2]` / `[Config HybridSDSM_v2]`
`maxObjectsPerSdsm`	16	32
`K_max`	32 (NED default)	32 (explicit in v2 configs)
`detectionRange`	300 m	(inherits)
`detectionMaxAge`	2 s	(inherits)

v2 parallel-threshold scheduler + Hybrid object pipeline (NED defaults)

Used when greedyVariant="v2" and/or hybridVariant="v2". Values below are defaults unless you uncomment overrides in omnetpp.ini under the v2 configs.

Group	Parameter	Default
Normalize	`refSelfChange`	10.0
	`refDist`	5.0
	`T_max`	5.0 s (backstop / time term scale)
	`K_max`	32 (object-set change denominator)
	`alpha_p` / `alpha_v` / `alpha_h`	1.0 / 0.5 / 2.0
Thresholds	`tau_self` / `tau_obj` / `tau_time` / `tau_cbr`	0.4 / 0.4 / 0.5 / 0.6
VoI (Lyu-style)	`w_novelty` / `w_quality`	0.5 / 0.5
	`tau_decay` / `tau_quality`	1.5 s / 2.0 s
	`d_ref`	50 m
Confidence (future)	`p_highConfidence` / `tau_conf`	0.85 / 0.5 (inactive at `conf=1.0` in code today)
Spatial association	`assocCoarseGate`	5 m
	`assocChiSquaredThreshold`	13.28
	`assocSigmaPosSquared` / `assocSigmaVelSquared`	1.0 / 0.25
	`assocPruneAge`	3 s
Redundancy (LARM-inspired)	`redundancyWindow`	0.5 s

Logging

Parameter	Default	Notes
`csvLoggingEnabled`	true	Master CSV switch
`txrxLogEnabled`	false	Combined TX/RX stream
`rxLogEveryNth`	1	e.g. 2 in `[Config EventTriggered]` to thin `rx.csv`
`logPrefix`	`"default"`	Per-`[Config]` → `Periodic`, `Greedy_v2`, etc.
`runNumber`	0	Set to `${runnumber}` in ini
`timeseriesSampleInterval`	1.0 s	`*-timeseries.csv` cadence
Object-AoI sampler interval	0.1 s	Hard-coded in `RosSDSMApp` (`objectAoiSampleInterval_`); not a NED parameter

Algorithms

Canonical v2 (primary comparison)

All three share the same PHY/MAC, SDSM schema, and (for the two adaptive policies) the same 100 ms evaluation tick.

Algorithm	When to send	What goes in the SDSM (objects)
`Periodic`	Fixed 10 Hz (`sendInterval = 0.1 s`)	Distance top‑K; payload capped by `maxObjectsPerSdsm` (16 under default `[General]`, unless you raise it for fair payload size)
`Greedy_v2`	`evaluateV2Schedule`: parallel thresholds on self-change, object-set change, time; OR combine; CBR suppressor; backstop at T_max; min inter-send (backstop can override)	Distance top‑K; `maxObjectsPerSdsm = 32` and `K_max = 32` in `[Config Greedy_v2]`
`HybridSDSM_v2`	Same scheduler code as `Greedy_v2` (reason prefix `hybrid_` vs `greedy_` in `*-triggers.csv`)	RX spatial association → LARM-style redundancy window → Lyu-style VoI → top‑K; `maxObjectsPerSdsm = 32`; confidence is `conf = 1.0` until a perception module supplies scores

Factor isolation

Periodic vs Greedy_v2: isolates scheduler (fixed vs adaptive).
Greedy_v2 vs HybridSDSM_v2: isolates object selection (same scheduler).

Scheduler logic is ETSI TS 103 324 / TS 102 687–inspired, not a conformance certification. See comments in src/RosSDSMApp.cc (evaluateV2Schedule).

Legacy v1 (preserved, not canonical for the current study)

Greedy, EventTriggered, GreedyBSMImplied, HybridSDSM (hybridVariant="v1") remain in simulations/omnetpp.ini for reproducibility. They use the older weighted-sum / v1 hybrid paths. Prefer v2 configs for new results.

SDSM payload (J3224-aligned)

Schema supports up to K_max objects (default 32 in NED); maxObjectsPerSdsm in omnetpp.ini is the packer cap (16 in [General], 32 in [Config Greedy_v2] / [Config HybridSDSM_v2]). SDSM_PER_OBJECT_BYTES = 26; obj_measurement_time_ms per object.
Objects are drawn from neighborInfo_ within detectionRange (default 300 m) and detectionMaxAge (default 2 s).

Metrics and CSVs

Summary (`*-summary.csv`)

Column	Meaning
`total_tx`, `total_rx`	Global SDSM send / successful receive counts
`avg_one_way_latency`, `p95_`, `p99_`	`simTime − sendTimestamp` at receiver (seconds). Not Kaul et al. sawtooth AoI.
`sender_state_redundancy_rate`	Fraction of RX where sender’s position+speed delta vs previous message < epsilon
`avg_throughput`	`(total_rx / sim_duration) / num_vehicles` — receptions per second per vehicle (not bytes/s)
`pdr_legacy_all_pairs`	`total_rx / (total_tx × (num_vehicles−1))` broadcast-style PDR (optimistic denominator); distance-binned PDR: `analysis/compute_pdr.py`
`avg_object_aoi`, `p95_`, `p99_`	v2 sampler: age of last update per tracked object; −1 if unused (e.g. Periodic). Absolute seconds can be inflated by stale tracks — compare across algorithms or condition on distance/relevance offline.
`assoc_*`	Spatial-association stage totals (HybridSDSM_v2; 0 for Greedy_v2 / Periodic)

Per-vehicle (`*-vehicle-summary.csv`)

Includes avg_latency / p95_latency from simTime − BSM envelope timestamp, and avg_one_way_latency / p95_one_way_latency from payload sendTimestamp (different clocks).

Reception log (`*-rx.csv`)

time,receiver,sender,message_id,one_way_latency,inter_arrival,snr,rss_dbm,distance_to_sender,packet_size,cbr,num_objects,delta_state,sender_state_redundant

v2-only logs

*-triggers.csv: one row per vehicle per tick (greedy_* / hybrid_* reasons).
*-object-aoi.csv: 100 ms samples (receiver, object_id, aoi).

Post-processing (repo)

Script	Purpose
`analysis/compute_aoi.py`	Kaul-style AoI from full `*-rx.csv`
`analysis/compute_pdr.py`	Distance-binned PDR
`analysis/compute_redundancy.py`	Object-level redundancy (v2 + object-AoI)
`analysis/scale_by_distance.py`	Auxiliary scaling / distance analysis helper

Limitations (disclose in papers/talks)

No urban obstacle shadowing — LOS-style links with fading only.
Hybrid confidence not driven by a real detector (conf = 1.0).
Association uses greedy one-to-one matching after Mahalanobis gating, not an optimal assignment solver.
ETSI-inspired scheduler/suppressor — not a standards compliance claim.
num_vehicles may be < requested N; always use metadata for fair normalization.
Single-scenario / seed until you publish multi-seed CIs.

Sources (design + standards)

SAE J3224 — SDSM structure.
SAE J2945/1 — periodic safety messaging context (10 Hz baseline).
ETSI TS 103 324 — parallel-threshold style motivation (scheduler).
ETSI TS 102 687 — DCC / CBR motivation.
S. Kaul, R. Yates, M. Gruteser — Age of Information (for compute_aoi.py definition).
T. Thandavarayan et al., JNCA 2023 — LARM (redundancy gate inspiration).
X. Lyu et al., IEEE VNC 2025 — VoI-style object ranking.
C. Sommer et al. — adaptive beaconing / self-change literature (v1 Greedy lineage).

ROS message alignment: ucla-mobility/CPX-SDSM.

Setup

Prerequisites: OMNeT++ 6.0+, Veins 5.2+, SUMO 1.8+. Set *.manager.commandLine in simulations/omnetpp.ini to your sumo binary.

source <omnetpp-install>/setenv
cd src && make -j$(nproc)

Running

Canonical v2 (example)

cd /path/to/veins_ros_v2v_ucla
python3 run_experiments.py -a Periodic      -s 0 -n 400 --skip-probe
python3 run_experiments.py -a Greedy_v2     -s 0 -n 400 --skip-probe
python3 run_experiments.py -a HybridSDSM_v2 -s 0 -n 400 --skip-probe

Artifacts: results/n400/<Algorithm>/seed0/ (and optional symlinks under results/n400_seed0_bundle/ if you use them).

Legacy v1 examples

python3 run_experiments.py --algorithm Greedy --sim-duration 90
python3 run_experiments.py --algorithm HybridSDSM --sim-duration 90

Shrink huge RX CSV (optional)

python3 scripts/shrink_rx_csv.py results/n400/Periodic/seed0/Periodic-r0-rx.csv --every 4 --decimals 3

ROS 2 bridge (optional)

cd ros2_ws && colcon build && source install/setup.bash
ros2 run veins_ros_bridge udp_bridge_node --ros-args -p udp_port:=50010

Enable with rosBridgeMode = "live" in omnetpp.ini when needed.

License

See upstream ucla-mobility/CPX-SDSM.

Name		Name	Last commit message	Last commit date
Latest commit History 7 Commits
analysis		analysis
ros2_ws/src		ros2_ws/src
scripts		scripts
simulations		simulations
src		src
sumo		sumo
.gitignore		.gitignore
.nedfolders		.nedfolders
Makefile		Makefile
README.md		README.md
REWRITE_SUMMARY.md		REWRITE_SUMMARY.md
claude.md		claude.md
run_all_non_ros.sh		run_all_non_ros.sh
run_density_sweep.py		run_density_sweep.py
run_experiments.py		run_experiments.py
run_in_terminal.sh		run_in_terminal.sh
setup_env.sh		setup_env.sh

Folders and files

Latest commit

History

Repository files navigation

veins_ros_v2v_ucla

Simulation stack

Simulation world (current defaults)

Implementation parameters (reference)

Experiment runner (run_experiments.py)

SUMO (sumo/scenario.sumo.cfg)

OMNeT++ network & TraCI (simulations/omnetpp.ini [General])

PHY analogue models (simulations/config.xml)

RosSDSMApp — mode switches (NED defaults; override in omnetpp.ini)

Greedy / hybrid v1 utility (NED + [General] in omnetpp.ini)

Multi-object SDSM caps & sensing window

v2 parallel-threshold scheduler + Hybrid object pipeline (NED defaults)

Logging

Algorithms

Canonical v2 (primary comparison)

Legacy v1 (preserved, not canonical for the current study)

SDSM payload (J3224-aligned)

Metrics and CSVs

Summary (*-summary.csv)

Per-vehicle (*-vehicle-summary.csv)

Reception log (*-rx.csv)

v2-only logs

Post-processing (repo)

Limitations (disclose in papers/talks)

Sources (design + standards)

Setup

Running

Canonical v2 (example)

Legacy v1 examples

Shrink huge RX CSV (optional)

ROS 2 bridge (optional)

License

About

Resources

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Experiment runner (`run_experiments.py`)

SUMO (`sumo/scenario.sumo.cfg`)

OMNeT++ network & TraCI (`simulations/omnetpp.ini` `[General]`)

PHY analogue models (`simulations/config.xml`)

`RosSDSMApp` — mode switches (NED defaults; override in `omnetpp.ini`)

Greedy / hybrid v1 utility (NED + `[General]` in `omnetpp.ini`)

Summary (`*-summary.csv`)

Per-vehicle (`*-vehicle-summary.csv`)

Reception log (`*-rx.csv`)

Packages