Where AntHocNet could go next — research landscape (2026-10)¶
Status: a web- and abstract-level survey from 2026-10-09, written for the post-v2.0.0 replan (roadmap.md). It shares the limits of satellite-routing-prior-art.md:
- each result below is quoted from an abstract or a summary, not checked against the full text;
- no number here may be cited in a publication until someone has read the paper;
- "no 2024–2025 paper found" means this search found none — it is not a claim that the literature has none.
The page exists so the replan's choices can be traced to sources, and so the next person to pick up a family starts from a reading list rather than a blank search box.
1. The question¶
v2.0.0 closed the family axis this project committed to:
- MANET (the grid);
- static mesh;
- FANET;
- VANET;
- satellite, both the ISL torus and the moving Walker / Starlink shell.
Each family has a results page with CIs, and there is a cross-family ranking statement. The question this page answers is the maintainer's (2026-10-09):
- what other networks could the protocol be evaluated on;
- which algorithm adjustments could improve specific cases;
- which research should inform both.
Two existing constraints bound every answer below:
- ADR-0019: a family changes the scenario, never the protocol defaults. A per-regime improvement must be a gated mechanism with its own default-off switch, earned by an A/B on identical seeds.
- ADR-0020: default-off extensions must be provably byte-identical when off. This is the template for every mechanism proposed in §3.
2. Candidate network families¶
The table assesses each candidate on four questions:
- does the literature use ACO there;
- is there an ns-3 substrate (ADR-0023 keeps one simulator);
- would AntHocNet's mechanisms actually be exercised;
- what is the risk.
| family | ACO / swarm literature | ns-3 substrate | what it would test | risk | verdict |
|---|---|---|---|---|---|
| Disaster / emergency response — partitioned first-responder teams, mixed indoor/outdoor shadowing | MANET protocol comparisons under disaster-area mobility find connectivity "varies widely enough to be hard for current routing protocols" (Raffelsberger & Hellwagner, WISES 2012); composite rescue-team mobility models exist (Reina et al.; RTTMM, Gondaliya & Atiquzzaman) | stock ns-3 (mobility model + shadowing are scenario work); reuses the open partition/merge stress item #62 | repair under partition and re-merge, which no current family exercises; group mobility | low — no new substrate | adopt (family axis III) |
| Space-air-ground integrated (SAGIN) — LEO shell + HAPS/UAV relay layer + ground | surveys flag routing across vertically heterogeneous layers as open: "the current routing protocol is not applicable to vertical space networks" (2020 survey, arXiv 2002.08811); HAPS as a central NTN component is under-studied (arXiv 2510.19731, 2025) | extends leo-walker (stock ns-3.48 LEO, ADR-0022) with an air layer: the FANET mobility already exists |
heterogeneous link delays (ms ISL vs sub-ms air), a third tier for the pheromone to choose between | medium — harness work, no new module | adopt (family axis III) |
| Tactical narrowband MANET — 9.6–64 kbit/s radios, high latency | no ACO work found; 2012 tactical-mobility study concludes no protocol "routes efficiently across all network sizes, loads, and mobility levels" (Kioumourtzis et al.); ns-3/CORE behaviours not seen on hardware (Grandhomme et al. 2016) | stock ns-3 (rate-limited PHY) | control overhead as the binding constraint — AntHocNet's NRL lead/lag matters most here | low substrate, high relevance-of-claim risk (no open benchmark to anchor to) | spike — one cell inside the disaster family, not a family of its own |
| Underwater acoustic (UASN) — ~1500 m/s propagation, long delays, energy-bound | ACAR (IET Comms 2020), PB-ACR (IEEE Access 2021) show ACO routing there, both energy-focused, evaluated on NS-2 Aqua-Sim | Aqua-Sim NG is a third-party ns-3 add-on that recommends ns-3.40 — not in this repo's 3.36–3.48 matrix; Aqua-Sim FG is a separate newer codebase | propagation-dominated timing — the same mismatch as the satellite ISL (#205), at seconds instead of milliseconds | high — third-party substrate, version pin conflict, a whole new PHY/MAC | research spike only; revisit if #205 produces a propagation-aware timing mechanism worth testing at the extreme |
| LoRa mesh — duty-cycle limited, multi-km links | 2024 ACM Computing Surveys review: reactive protocols edge out proactive ones on scalability and power; hybrid only helps in specific topologies; ns-3 LoRaMesh raised far-node PDR 40.2 → 73.8 % (Sensors, 2025) | ns-3 lorawan module (third party) |
whether ants fit inside a 1 % duty cycle at all | high — ant overhead may simply not fit the duty cycle | out (as WSN/IoT, RPL's problem); recorded so it is not re-proposed without new evidence |
| Maritime (ship ad hoc, VHF) — sparse, long-range, partitioned | SANET studies compare DSDV/AODV/AOMDV/DSR; MADNET switches MANET↔DTN by connectivity | stock ns-3 | sparse, long-range partitions | medium | out for now — its distinctive need (DTN store-carry-forward) is a non-goal; reopens with DTN below |
| mmWave / directional mesh (UAV swarms) | directional FANET routing survey (2021); 2024 survey on neighbour discovery / beam alignment in mmWave UAV swarms (arXiv 2410.11490) | ns-3 mmWave modules (third party, heavy) | beam alignment breaks the broadcast assumption hellos and reactive floods rely on | high — breaks a protocol assumption, not a scenario knob | out — it is a protocol redesign, not a family |
3. Algorithm adjustments worth measuring (all gated, default-off)¶
These are mechanisms, not presets. Each one must:
- (a) have a default-off attribute;
- (b) pass a byte-identical determinism check when off;
- (c) be A/B'd on identical seeds in the family it targets and in every other family, so that a regression elsewhere is visible;
- (d) update the mechanism × regime table in network-regimes.md.
| mechanism | measured weak spot it targets | literature | existing issue |
|---|---|---|---|
| Link-lifetime prediction — discount pheromone on a next hop whose predicted link expiry is near (position + velocity) | VANET: 36 % of AntHocNet's traffic lost to reconvergence; oracle 81.7 % vs AntHocNet 42.1 % — the widest gap of any family (#537) | intersection-aware link lifetime (iCAR, 2013); ML link-lifetime prediction (Sensors 2022, doi:10.3390/s22166038); ACO + link prediction (ERIACO, 2024); mobility-anticipated ETX gives PDR close to 1 in ns-3 (HAL hal-01072234) | #574 (needs the position port #573) |
| SINR link metric | fidelity: every thesis headline result uses it, and this repo's default is the metric the thesis benchmarks as worse | Neishaboori & Kesidis 2008 (SINR as secondary metric on ETX, damps instability); ETX/ETT comparisons (Draves et al., Microsoft Research) | #181, through the #142 seam |
| Propagation-dominated timing (T_hop, hello, lifeAnt, repair waits) | S1: AntHocNet 15.5 pp under the delay oracle (7.0 pp on walker16); the gap grows with path length | — (this repo's own finding, ADR-0019 cites it as the example) | #205 (v2.3.0) |
| Hello suppression on point-to-point ISLs | satellite: hellos are redundant on a link with one known peer (network-regimes §6) | — | #204 |
| Quiet mode for stable topologies — proactive-ant back-off when sampled routes stop changing | static mesh: OLSR leads (99.82 % vs 99.34 % PDR, NRL 1.78 vs 4.51); ants keep sampling links that never change (the learn site's mesh challenge shows it) | adaptive evaporation for dynamic optimisation (Mavrovouniotis & Yang 2013/2014); counterpoint: Pellegrini, Stützle & Birattari 2012 on when parameter adaptation helps | #571 |
| Adaptive evaporation — evaporation rate tracks observed route churn | handover-heavy shells (walker16 hop changes 2.07 vs the oracles' ~1.0 per flow-minute) and VANET corners | Mavrovouniotis & Yang (EvoApplications 2013; IEEE CIDUE 2014): self-adaptive evaporation beats fixed rates on dynamic problems — tested on dynamic TSP/VRP, not packet routing, so transfer needs its own validation | #572 |
| Energy-aware link metric | FANET energy per delivered bit is published (#508) but nothing optimises it | BeeAdHoc (GECCO 2005: energy savings from fewer control packets + multipath); PEEBR (predicted residual battery); min-energy vs max-min residual routing | #145 |
| Re-injection that tells redundant from delivering | 65–67 % duplicate rate by direct measurement (reinjection.md) | — | #430 |
| RepairHoldCap | ~168 ms of tail the #371 flip left on the table | — | #433 |
Note on position-aware mechanisms. Link-lifetime prediction (and any
geographic hint) needs node position and velocity inside core/. Today core/
sees neither. That is a new port in the ADR-0003 sense — the adapter supplies
it, the core never reads a simulator. It must be absent (not zero) when an
adapter cannot supply it, so the default path stays byte-identical. This
deserves its own ADR before code.
4. Comparators the literature now expects¶
- Other swarm protocols. These are the natural "is it the ants, or this design of ants?" control. As of this search, none has a maintained ns-3 implementation; past comparisons ran on NS-2 or OMNeT++, pairwise, in scenarios that cannot be lined up. No study was found that compares them all under identical conditions.
| protocol | design | what it tests against AntHocNet | plan |
|---|---|---|---|
| ARA (Güneş et al. 2002) | purely reactive ACO, built to cut overhead (roots in ABC and AntNet) | whether proactive sampling and repair pay for themselves | v2.5.0 · #586 |
| Termite (Roth & Wicker, SIDM 2005) | stigmergy: routing information rides inside data packets, no control ants; randomised multipath | a different overhead model; its authors report it beats AODV on primary metrics | v2.5.0 · #587 |
| BeeAdHoc (Wedde et al., GECCO 2005) | bee-inspired source routing, scouts and foragers; low energy from fewer control packets | a second swarm family, and the energy-per-bit metric | stretch · #589 |
| HOPNET (Wang, 2007 thesis) | ants hopping between routing zones | the only result found benchmarking against AntHocNet directly; it claims better scaling | stretch |
| AntNet (Di Caro & Dorigo 1998) | ACO for wired networks | — | skip: AntHocNet's ancestor, not a competitor |
The risk is a strawman: a quickly written competitor loses for the wrong reason. This repository has paid for that once (#425/#416: two vendored arms compiled, passed CI and forwarded nothing). Each arm therefore needs a fidelity sheet from its paper, the per-PR delivery smoke (#439), and an anchor reproducing its own paper's headline trend against AODV before its numbers are published. An NS-2 thesis from Thapar University that compared an ant scheme with ARA and AntHocNet is the closest earlier attempt found.
- Learned routing (DRL / MARL). This is the most-cited new comparator family:
- DeepCQ+ (Kaviani et al., arXiv 2101.03273) reports 10–15 % over Q-routing and robustness outside the training range;
- Alanazi & Zareei (IEEE Access 2025) pair MADRL with GNNs;
- a 2024 FANET routing review (Alexandria Eng. J., doi:10.1016/j.aej.2024.09.032) builds its taxonomy around RL;
- tooling exists (ns3-gym, Gawłowicz & Zubow, MSWiM 2019; PRISMA, a multi-agent RL routing playground on ns-3).
No head-to-head reproducible benchmark was found. That is both the risk (no agreed setup) and the opportunity (this repo's 20-seed paired-test harness is exactly what that comparison lacks). The roadmap's existing non-goal says a DRL baseline is "planned but gated: a leaky comparison would damage credibility". The replan keeps the gate and states what lifts it (train/test seed split, held-out families, the training budget reported). - Q-routing. A tabular, non-deep RL comparator is cheap, old (Boyan & Littman 1994) and the natural "learning but not ants" control. - Hybrid ACO variants. In the 2024–2025 literature found, recent ACO-for-LEO work mostly hybridises ACO with another metaheuristic to escape stagnation: - ACO + sparrow search (IEEE ICDSCA 2024); - SAT-IACO (Springer 2025), for satellite IoT access rather than routing.
None reports an ns-3 or comparable-harness evaluation in what was found.
4b. Secure routing comparators (after v3.0.0)¶
The v3.0.0 security epic (#302): - scopes blackhole, grayhole, pheromone forgery and replay; - leaves wormhole, Sybil and rushing out; - compares only against unprotected AODV, OLSR and DSDV.
Showing that a defence is competitive needs secure comparators.
| protocol | approach | base protocol (ns-3 module here) | plan |
|---|---|---|---|
| SAODV | digital signatures + hash chains on AODV control messages | AODV ✓ | v3.1.0 · #592 |
| SEAD | hash-chain authenticated distance vector | DSDV ✓ | v3.1.0 · #593 |
| TAODV | per-neighbour trust, no cryptography | AODV ✓ | v3.1.0 (the trust side of the cryptographic-vs-trust axis) · #594 |
| Ariadne | authenticated source routing (TESLA) | DSR (stock ns-3) | v3.1.0 stretch |
| ARAN | certificate-signed routing | AODV-like | v3.1.0 stretch |
| BeeSec / BeeAIS / BeeAIS-DC (Mazhar & Farooq 2007) | asymmetric-key or artificial-immune-system security on BeeAdHoc | BeeAdHoc (v2.5.0 stretch → prerequisite) | v3.2.0 · #595 |
| Trust-weighted ACO (Simaremare et al., ICC 2014) | ants deposit positive pheromone only through trusted nodes | ARA (v2.5.0) + trust | v3.2.0 · #596 |
| ACO + watchdog (Kalinin et al. 2018) | every node an agent that rates its neighbours' security | ARA + watchdog | v3.2.0 · #597 |
What the search found: - Simulators. Comparative studies of SAODV, Ariadne, SEAD and ARAN ran on NS-2, GloMoSim or real hardware, always pairwise or in small groups. No ns-3 study was found, and no study puts all of them under one attack model. - Cost of security. Security costs performance; the SEAD study says so directly. A hardware comparison of SAODV against TAODV (Stevens) also notes that cryptographic schemes open new denial-of-service avenues. - Mobility model. One group argues that earlier secure-routing evaluations relied on random waypoint, which does not converge at high pause times. This repo's mobility and warm-up policy already address that. - Swarm literature. It is mostly smaller venues with self-reported simulations (for example, 99.66 % PDR claimed for trust-ACO AODV). The anchor gate exists for exactly this. - Gap. No evaluation was found of pheromone poisoning: colluding nodes forging or inflating trails. This attack is specific to ant routing, and it is v3.3.0's original contribution. - Crypto in simulation. ns-3 does not execute cryptography. The standard practice, and the plan here, is to model each signature or verification as a computation delay, measured on stated hardware, plus its bytes on the wire.
5. Store-carry-forward (DTN) — why it stays out, and what would bring it in¶
DTN routing is a capability AntHocNet structurally lacks: it drops a packet it
cannot forward after QueueTimeout. The swarm literature has ACO-for-DTN
designs:
- ACR (Yang et al., 2012): ≥ 25.8 % lower delay than other forwarding schemes on the Infocom/RollerNet traces;
- GrAnt (SBrT): more deliveries with fewer replicas than Epidemic and PROPHET.
What would bring it in: suppose the disaster family (§2) shows that partition-bound loss dominates AntHocNet's shortfall there, i.e. the oracle also fails, as the sparse cells already show. Then an opt-in carry buffer is the measured next step, and maritime (§2) comes with it.
6. Sources¶
Every entry was retrieved on 2026-10-09 at abstract or summary level.
Families - Raffelsberger & Hellwagner, Evaluation of MANET routing protocols in a realistic emergency response scenario, WISES 2012 — CCS Labs entry, PDF - Reina et al., Ad hoc network in a disaster area: a composite mobility model and its evaluation — HAL - Kioumourtzis et al., tactical MANET routing evaluation (2012) — Univ. Patras - Grandhomme et al., Comparison of inter-MANET routing protocol evaluation tools (2016) — EURECOM - Satellite Communications in the New Space Era (survey) — arXiv 2002.08811 - Bridging Earth and Space: A Survey on HAPS for Non-Terrestrial Networks (2025) — arXiv 2510.19731 - Survey on Near-Space Information Networks — arXiv 2310.09025 - ACAR (ant colony routing for UASN), IET Communications — IET - Aqua-Sim NG (ns-3 port) — GitHub; Aqua-Sim FG — arXiv 2410.20698 - Durand & Booysen, Performance Evaluation of a Mesh-Topology LoRa Network, Sensors 2025 — PMC - Multi-hop and mesh for LoRa networks (survey) — UMS eprints - Maritime communication review, J. Mar. Sci. Eng. 12:1264 (2024) — MDPI - Survey on Neighbor Discovery and Beam Alignment in mmWave-Enabled UAV Swarm Networks (2024) — arXiv 2410.11490
Mechanisms - Mobility-anticipated ETX — HAL hal-01072234 - Neishaboori & Kesidis, SINR-sensitive routing in wireless 802.11 mesh networks — Penn State - Draves, Padhye & Zill, mesh routing metrics — Microsoft Research - iCAR: intersection-based connectivity-aware routing — ICESI repository - ML link-lifetime prediction in VANETs, Sensors 2022 — doi:10.3390/s22166038 - Driving path stability in VANETs — arXiv 1906.08370 - Mavrovouniotis & Yang, adaptive / self-adaptive evaporation — CUT repository, Springer - BeeAdHoc, GECCO 2005 — PDF; PEEBR — SAI - Hybrid ant-colony inter-cluster routing for FANET, Sci. Rep. 2024 — PMC
Comparators - DeepCQ+ — arXiv 2101.03273 - ns3-gym — arXiv 1810.03943 - ACO + sparrow search for LEO routing (2024) — BUAA - SAT-IACO (2025) — Springer - Zhang et al., ACO-based MANET routing survey, IEEE Access 2017 — IEEE Xplore
Swarm comparators - Güneş et al., ARA — the ant-colony based routing algorithm for MANETs (2002) — FU Berlin - Roth & Wicker, Termite (SIDM 2005) — PDF; Roth thesis — Cornell eCommons; seminar critique — Freiburg - Wang, HOPNET (2007 thesis) — UManitoba MSpace - Ducatelle, AntHocNet thesis (2007) — USI - Performance Analysis of Swarm Based Routing Protocols for MANETs (NS-2 thesis, Thapar University) — TUDR
Secure comparators - Sadasivam, Changrani & Yang, SEAD vs DSDV vs DSR — UHCL - Evaluations of Secure MANET Routing Protocols (Ariadne, SAODV under attack; thesis) — UHCL - ARAN vs AODV under blackhole and IP spoofing (GloMoSim) — CUP - Cryptographic versus trust-based methods for MANET routing security (SAODV vs TAODV on hardware) — Stevens - Secure bee algorithms for MANET routing (survey; BeeSec, BeeAIS) — UB - Secure routing in MANETs: a bio-inspired approach with honey bees — Inderscience - Kalinin et al., attack prevention in self-organizing ad hoc networks using swarm intelligence (2018) — doi:10.3103/S0146411618080163 - Trust-based ant routing under DoS/DDoS (ICC 2014) — UI - Blackhole prevention in MANET using ACO — ITC - ACO approach for wormhole detection in MANETs (2023) — Inderscience - Trust system and multiple verification for wormholes, IEEE Access 2024 — Sogang
DTN - Routing in delay-tolerant networking (overview) — Wikipedia - ACR, ant-colony DTN routing (2012) — CRAD - GrAnt — SBrT