MeshCore

2980 Commits

6 Branches

98 Tags

234 MiB

Tree: 11ca58a7aa

Commit Graph

Author	SHA1	Message	Date
Friday	d9e0d42738	sim: add ADAPTIVE routing strategy, density estimator, and TX power saving Adds a complete simulation harness for the MeshCore radio platform, including: ### ADAPTIVE routing strategy (sim/src/RoutingStrategies.h) Hash-based relay gate selects a deterministic subset of nodes to forward each flood, with relay percentage adapted to local neighbour density: - SPARSE (<=4 neighbours): 100% relay -- no redundancy to spare - MEDIUM (5-14): 25% relay - DENSE (>=15): 15% relay Selection uses hash(packet_seed XOR node_seed) with no coordination required. Different packets select different subsets for stochastic load balancing. Relay suppression: nodes cancel queued outbound TX if they overhear another node already relaying the same flood (matched by packet hash). ### DensityEstimator (sim/src/DensityEstimator.h) Passive sliding-window neighbour count from normal traffic -- no extra messages. Only counts direct senders (hop_count==1) over a configurable window (default 60s). Uses std::unordered_set for unique counting; no fixed-array cap. ### Adaptive TX power saving (sim/src/SimNode.h) power_save_enabled reduces TX power in DENSE/MEDIUM tiers: DENSE: configurable (default 10 dBm, saves 75% TX current vs 20 dBm) MEDIUM: configurable (default 14 dBm, saves 60% TX current) SPARSE: always full power -- every hop counts on marginal links Validated: 100% delivery maintained at -10dBm DENSE reduction. Energy savings ~35% radio total. TX power reduction also lowers area RF noise floor. ### Scenario suite (sim/scenarios/) scenario_adaptive: ADAPTIVE vs DEFAULT vs PATH_SNR_HYBRID comparison scenario_concurrent: 2-8 simultaneous flood sources scenario_longchain: 20-hop chains at marginal SNR scenario_mixed: ADAPTIVE + legacy DEFAULT interoperability scenario_dutycycle: EU 1% duty cycle enforcement validation scenario_relay_pct_sweep: grid sweep to find optimal DENSE%/MEDIUM% operating point scenario_txpower: TX power saving vs delivery rate vs energy trade-off ### Key empirical results FM50 ADAPTIVE vs DEFAULT: equal delivery rate, -65% airtime, -70% collisions CH20 (chain): ADAPTIVE = SPARSE throughout, identical to DEFAULT TX power MODERATE (-10dB): 100% delivery, -35% energy across all topologies Legacy interop: zero delivery regression with mixed firmware ### Bug fixes - DensityEstimator: replace seen[64] array with unordered_set (was capping at 64 neighbours) - SimBus::onTransmit: add len > 255 guard before memcpy into 255-byte buffer - SimNode p_forward gate: >= -> > (was silently dropping relays at p_forward=1.0) - SimNode TX power reset: remove redundant !power_save_enabled clause - scenario_concurrent: fix stale ConcurrentResult* pointer after vector reallocation - scenario_relay_pct_sweep: update TxCallback from 4-arg to 5-arg (tx_power_dbm) - scenario_adaptive: remove unused variable last_suppressed Co-Authored-By: Claude Sonnet 4.6 <[email protected]>	1 month ago

Author

SHA1

Message

Date

Friday

d9e0d42738

sim: add ADAPTIVE routing strategy, density estimator, and TX power saving

Adds a complete simulation harness for the MeshCore radio platform, including:

### ADAPTIVE routing strategy (sim/src/RoutingStrategies.h)
Hash-based relay gate selects a deterministic subset of nodes to forward each
flood, with relay percentage adapted to local neighbour density:
  - SPARSE  (<=4 neighbours): 100% relay -- no redundancy to spare
  - MEDIUM  (5-14):           25% relay
  - DENSE   (>=15):           15% relay

Selection uses hash(packet_seed XOR node_seed) with no coordination required.
Different packets select different subsets for stochastic load balancing.

Relay suppression: nodes cancel queued outbound TX if they overhear another
node already relaying the same flood (matched by packet hash).

### DensityEstimator (sim/src/DensityEstimator.h)
Passive sliding-window neighbour count from normal traffic -- no extra messages.
Only counts direct senders (hop_count==1) over a configurable window (default 60s).
Uses std::unordered_set for unique counting; no fixed-array cap.

### Adaptive TX power saving (sim/src/SimNode.h)
power_save_enabled reduces TX power in DENSE/MEDIUM tiers:
  DENSE:  configurable (default 10 dBm, saves 75% TX current vs 20 dBm)
  MEDIUM: configurable (default 14 dBm, saves 60% TX current)
  SPARSE: always full power -- every hop counts on marginal links

Validated: 100% delivery maintained at -10dBm DENSE reduction. Energy savings
~35% radio total. TX power reduction also lowers area RF noise floor.

### Scenario suite (sim/scenarios/)
  scenario_adaptive:        ADAPTIVE vs DEFAULT vs PATH_SNR_HYBRID comparison
  scenario_concurrent:      2-8 simultaneous flood sources
  scenario_longchain:       20-hop chains at marginal SNR
  scenario_mixed:           ADAPTIVE + legacy DEFAULT interoperability
  scenario_dutycycle:       EU 1% duty cycle enforcement validation
  scenario_relay_pct_sweep: grid sweep to find optimal DENSE%/MEDIUM% operating point
  scenario_txpower:         TX power saving vs delivery rate vs energy trade-off

### Key empirical results
  FM50 ADAPTIVE vs DEFAULT:  equal delivery rate, -65% airtime, -70% collisions
  CH20 (chain):              ADAPTIVE = SPARSE throughout, identical to DEFAULT
  TX power MODERATE (-10dB): 100% delivery, -35% energy across all topologies
  Legacy interop:            zero delivery regression with mixed firmware

### Bug fixes
  - DensityEstimator: replace seen[64] array with unordered_set (was capping at 64 neighbours)
  - SimBus::onTransmit: add len > 255 guard before memcpy into 255-byte buffer
  - SimNode p_forward gate: >= -> > (was silently dropping relays at p_forward=1.0)
  - SimNode TX power reset: remove redundant !power_save_enabled clause
  - scenario_concurrent: fix stale ConcurrentResult* pointer after vector reallocation
  - scenario_relay_pct_sweep: update TxCallback from 4-arg to 5-arg (tx_power_dbm)
  - scenario_adaptive: remove unused variable last_suppressed

Co-Authored-By: Claude Sonnet 4.6 <[email protected]>

1 month ago

1 Commits (11ca58a7aaafbf1396173424c615557562152f4f)