MAPC: Multi-Access Point Coordination
- MAPC is a coordination framework that synchronizes multiple APs by sharing TXOPs, achieving improvements like at least 25% higher throughput and reduced latency in dense WLAN deployments.
- MAPC employs modes such as Co-TDMA, Co-SR, and Co-BF where explicit control signaling orchestrates sequential or simultaneous transmissions to optimize airtime and manage interference.
- Advanced analytical and learning-based approaches, including CTMC models and RL algorithms like DQN and PPO, are used to dynamically schedule and adapt MAPC for robust network performance.
Multi-Access Point Coordination (MAPC) is a set of architectures and MAC/PHY mechanisms in which multiple access points coordinate time, frequency, and/or spatial resources instead of operating as independent CSMA/CA nodes. In IEEE 802.11bn, MAPC is a central mechanism for Ultra High Reliability (UHR): one AP that wins channel access can share its TXOP with other coordinated APs and, in some cases, enable simultaneous transmissions. The current 802.11bn tutorial places MAPC at the core of the amendment’s UHR agenda, including at least 25% improvement in throughput, at least 25% reduction in 95th-percentile MAC PDU latency, and at least 25% reduction in MPDU loss relative to 802.11be under comparable conditions, while the broader survey literature situates MAP-Co within the longer evolution from 802.11ax spatial reuse and 802.11be multi-AP proposals toward tightly coordinated multi-BSS operation (Wilhelmi et al., 11 Jun 2026, Wilhelmi et al., 24 Jun 2026, Verma et al., 2023).
1. Standardization setting and conceptual scope
MAPC in current WLAN research is primarily associated with IEEE 802.11bn, but its lineage is broader. The survey literature treats MAP-Co as a key feature of emerging WLANs, especially IEEE 802.11be and future 802.11ay-class systems, where dense OBSS deployments, stringent latency targets, and increasing inter-AP interference make purely distributed access progressively less effective (Verma et al., 2023). Within 802.11bn itself, MAPC is not a single procedure but a framework that encompasses coordinated time-domain, frequency-domain, and spatial-domain operation.
A recurring misconception is to equate MAPC solely with CoMP-like joint transmission. The 802.11bn tutorial and simulator literature instead present MAPC as an umbrella that includes Coordinated TDMA, Coordinated Spatial Reuse, Coordinated Beamforming, Coordinated OFDMA, and coordinated Non-Primary Channel Access, while also identifying Joint Transmission as a direction beyond 802.11bn rather than as a baseline feature (Wilhelmi et al., 11 Jun 2026, Wilhelmi et al., 24 Jun 2026). This broader framing matters because the standard’s immediate objective is not only peak throughput, but also deterministic access, lower collision probability, bounded interference, and better tail-latency behavior under dense multi-AP coexistence.
Architecturally, the survey distinguishes master-controller-based and master-AP-based realizations, and it separates loose coordination from tight coordination according to how much CSI, traffic, interference, and scheduling information is exchanged (Verma et al., 2023). This suggests that MAPC should be understood less as a single protocol and more as a coordination layer whose implementation range spans from AP-centric TXOP sharing to explicitly centralized, backhaul-assisted scheduling.
2. Coordination model, entities, and signaling primitives
In the 802.11bn notion adopted by Kom8ndor, MAPC “allows a coordinating AP (the AP that wins the channel access) to share its TXOPs with other coordinated APs and, in some cases, perform simultaneous transmissions” (Wilhelmi et al., 24 Jun 2026). Three entities are fundamental. The coordinating AP is the AP that wins channel access via EDCA/DCF or another supported access method. The coordinated APs are APs with an established MAPC agreement with the coordinating AP. The shared TXOP is the airtime interval inside which the coordinating AP may allocate time slices, invoke synchronized parallel transmissions, or apply coordinated beamforming.
This coordination is realized through an explicit control handshake. The AP that wins access sends an Initial Control Frame (ICF); coordinated APs respond with Initial Control Response (ICR); the coordinating AP may then send a Trigger Frame (TF) to synchronize multi-AP PPDUs; and Co-TDMA specifically uses MU-RTS to schedule time slots (Wilhelmi et al., 24 Jun 2026). In Kom8ndor, these operations are reflected in the per-node FSM through states such as TRANSMIT_ICF, WAIT_ICR, and WAIT_MU_RTS, while non-participating STAs set their NAV to the end of the coordinated TXOP and remain silent for its duration.
The standard mechanism is AP-centric rather than controller-centric: Co-TDMA itself is logically centralized at the coordinating AP, and there is no mandatory external network controller in the procedure. At the same time, research simulators and survey architectures routinely add a higher-level controller for orchestration, group formation, ML policy distribution, or multi-band coordination (Wilhelmi et al., 24 Jun 2026, Verma et al., 2023). A plausible implication is that standardized MAPC primitives are intentionally narrow, while deployment architectures may wrap them inside more centralized enterprise or industrial control planes.
3. Canonical MAPC modes
The simulator and tutorial literature identify three per-TXOP MAPC schemes as the main operational core of current 802.11bn studies: Co-TDMA, Co-SR, and Co-BF (Wilhelmi et al., 24 Jun 2026, Wilhelmi et al., 11 Jun 2026). They differ in how the coordinating AP uses the shared TXOP.
| Scheme | Principle | Salient mechanism |
|---|---|---|
| Co-TDMA | Sequential use of a shared TXOP | ICF/ICR plus MU-RTS schedule orthogonal time slices |
| Co-SR | Simultaneous transmissions with power coordination | ICF/ICR plus TF synchronize overlapping DL A-MPDUs |
| Co-BF | Simultaneous transmissions with beamforming and nulling | Shared TXOP plus coordinated ZF-based beams |
Co-TDMA divides the TXOP into orthogonal time slots assigned to participating APs. Kom8ndor’s default policy is equal airtime sharing, and the per-AP allowed data time depends on the AP’s requested duration , the standard maximum TXOP , the number of participating APs, and a common MAPC overhead term that accounts for ICF, ICR, MU-RTS, and SIFS intervals (Wilhelmi et al., 24 Jun 2026). Its main contribution is deterministic access and collision avoidance within the coordinated TXOP rather than raw spatial multiplexing gain.
Co-SR keeps transmissions simultaneous and coordinates transmit power instead of orthogonalizing airtime. In the first Kom8ndor release, the coordinating AP uses the transmit power configured in input_nodes, while coordinated APs use powers specified in the mapc file; the paper explicitly notes that smarter power assignment based on maximum allowable interference at each communication end is a natural extension, but not part of the present implementation (Wilhelmi et al., 24 Jun 2026). Conceptually, Co-SR sits between uncoordinated 802.11ax SR and full coordinated beamforming: it retains relatively low PHY complexity while making spatial reuse explicit and synchronized.
Co-BF is the most PHY-intensive current MAPC mode. Kom8ndor models each AP as a ULA and uses a Zero-Forcing construction so that each AP steers a beam toward its own STA and places nulls toward peer-BSS STAs. The steering vector is
and the ZF precoder is
which enforces unit gain in the desired direction and exact nulls in the constrained directions under the simulator’s perfect angular CSI assumption (Wilhelmi et al., 24 Jun 2026). Beamforming gain is applied only to data PPDUs; control frames remain omnidirectional.
These three schemes already show that MAPC is not reducible to “multi-AP transmission.” One branch pursues deterministic airtime partitioning, another coordinated interference budgeting, and another interference suppression via beam synthesis. The standard framework is therefore a coordination substrate onto which multiple MAC/PHY philosophies can be mapped.
4. Analytical and optimization frameworks
Analytical work on MAPC has focused especially on Coordinated Spatial Reuse. A CTMC-based throughput analysis for IEEE 802.11bn C-SR models its throughput and spatial efficiency and reports average throughput gains of up to 59% over legacy 802.11 DCF and up to 42% over 802.11ax OBSS/PD in the studied topologies (Wilhelmi et al., 2023). This positions C-SR as a coordinated alternative to purely threshold-based spatial reuse, with explicit multi-AP synchronization and interference control rather than opportunistic local aggressiveness.
Optimization-oriented work makes the scheduling problem more explicit. One recent formulation treats C-SR scheduling as the problem of selecting which devices transmit concurrently and with what settings, and derives a theoretical upper bound via mixed-integer linear programming optimized either for throughput or fairness. The same work then proposes flat MAB and hierarchical MAB schedulers, with H-MAB improving aggregate throughput over legacy IEEE 802.11 by 80% on average in random scenarios without reducing the number of transmission opportunities per station (Wojnar et al., 12 May 2025). In this line of work, MAPC becomes a combinatorial scheduling problem over feasible concurrent sets and power settings rather than only a procedural extension of EDCA.
A broader optimization strand extends MAPC beyond fixed AP geometry. In a 6DMA-enhanced coordinated uplink, the weighted sum rate is maximized by jointly optimizing the antenna position vector, the antenna orientation matrix, and the receive combining matrix over all coordinated APs under local antenna movement constraints. The resulting non-convex problem is handled through a Lagrangian-dual-style transform, alternating optimization, SCA for position updates, and Riemannian manifold optimization for orientation updates (Pi et al., 2024). This suggests that future MAPC research may not stop at time, frequency, and beams, but may also treat physical antenna geometry as a coordination variable.
5. Learning-based MAPC control
MAPC has become a natural target for online learning because coordination decisions are highly state-dependent and overhead-sensitive. Kom8ndor explicitly includes a machine-learning wrapper, supports decentralized learning, coordinated learning, and centralized orchestration, and exposes MAPC parameters such as scheme selection, Co-SR power levels, Co-TDMA slot shares, and group formation to MAB-style or external Python-based models, including DQN via a simple socket protocol (Wilhelmi et al., 24 Jun 2026). In this view, MAPC is not only a standardized feature set but also a control surface for adaptive policies.
A coordinated MA-MAB formulation for spatial reuse uses MAPC as a reward-sharing framework in which multiple agents tune SR parameters jointly. In Komondor-based evaluations, coordinated bandits increased mean throughput by 15%, improved minimum throughput across the network by 210%, and kept the maximum access delay below 3 ms (Wilhelmi et al., 2024). The technical significance is that MAPC here acts less as per-packet centralized scheduling and more as a coordination substrate for shared rewards and coupled exploration.
Deep RL pushes this further toward queue-aware, latency-oriented scheduling. A PPO-based scheduler trained in an 802.11bn-compatible Gymnasium environment uses queue states, delay metrics, and channel conditions to select multiple AP-STA pairs from SR groups, and reports up to 30% lower 99th-percentile delay than the best heuristic baseline across the studied loads and traffic patterns (Nunez et al., 25 Jul 2025). A related adversarial-RL formulation, originally motivated by multi-AP coordination against uncoordinated interference, shows that robustness to non-cooperating APs can improve the minimum sum of throughputs relative to training that ignores such interference (Kihira et al., 2020). Together, these results make clear that MAPC is increasingly studied as a sequential decision problem rather than only as a static feature catalogue.
6. Performance, implementation constraints, and future directions
The empirical MAPC picture is deliberately non-monotonic. In Kom8ndor’s tutorial 2-BSS example, Co-TDMA achieves slightly lower throughput than DCF because coordination overhead consumes airtime, even though it yields more predictable channel access and eliminates collisions within the coordinated TXOP. In the same setup, Co-SR outperforms both DCF and Co-TDMA, and Co-BF delivers the highest throughput because nulling raises SINR and supports simultaneous transmissions with limited mutual interference (Wilhelmi et al., 24 Jun 2026). MAPC gains therefore depend on network density, geometry, and overhead regime; stronger coordination is not automatically better.
CSI overhead is a particularly important constraint for Co-BF. A recent standards-aligned 802.11bn study shows that standard IEEE 802.11 CSI compression can make Co-BF underperform legacy transmissions in some situations because channel sounding overhead approaches the entire TXOP. An autoencoder-based CSI compression mechanism reduces channel sounding overhead by more than 50%, identifies a compression ratio of $1/4$ as the best accuracy/feedback-size tradeoff for lowest data latency, and in a 6-STA-per-AP case raises throughput by 39.9% over a legacy 40 MHz baseline while improving 99th-percentile latency (Aboushehada et al., 15 Apr 2026). This makes explicit a central implementation tension in MAPC: coordination quality can be limited less by the coordinated transmission itself than by the control and feedback machinery required to make it possible.
Several standard and research limitations remain open. The current draft restricts MAPC agreements to at most 2 APs, whereas Kom8ndor allows more than 2 APs per group in order to study “Wi-Fi 8 and beyond” behavior (Wilhelmi et al., 24 Jun 2026). The tutorial literature also points beyond the first MAPC wave toward interactions with MLO, OFDMA, R-TWT, Co-R-TWT, Co-CR, and eventually Joint Transmission (Wilhelmi et al., 11 Jun 2026). The broader survey on MAP-Co adds the unresolved burdens of CSI acquisition and sharing overhead, synchronization for CBF and JTX, scalability in dense deployments, legacy compatibility, and the coordination-gain-versus-overhead trade-off (Verma et al., 2023). A plausible implication is that MAPC will remain a layered research area: standard primitives may stabilize early, while practical performance will depend on higher-level scheduling, learning, compression, and clustering strategies that continue to evolve beyond the baseline amendment.