Coordinated TDMA (Co-TDMA): Principles & Applications

• Co-TDMA is a scheduling mechanism that explicitly assigns conflict-free time slots, reducing latency and jitter as shown in IEEE 802.11bn and vehicular networks.
• It uses centralized, distributed, and hybrid architectures to optimize slot allocation through methods such as TXOP sharing, state exchange, and multicoloring.
• Empirical results indicate that combining Co-TDMA with spatial reuse or redundancy significantly improves throughput, reliability, and performance in diverse wireless systems.

Coordinated TDMA (Co-TDMA) denotes a class of medium-access mechanisms in which transmissions are explicitly organized in time so that interfering entities use conflict-free transmission opportunities rather than relying solely on contention. In the IEEE 802.11bn literature, Co-TDMA is a multiple access point coordination (MAPC) technique in which an access point (AP) shares a portion of its obtained transmit opportunity (TXOP) with another AP; in system-level simulations, this reduces worst-case latency and jitter for low-latency traffic (Lee et al., 26 Aug 2025). More broadly, coordinated TDMA principles appear in centrally managed multi-hop wireless sensor networks, request/scheduling splits for vehicular safety messaging, and distributed TDMA-like schedules formed through reservation, state exchange, or local graph algorithms (Terraneo et al., 2018, Peng et al., 2019, Hui et al., 2011, 0902.1868).

1. Terminology and conceptual scope

In contemporary WLAN work, Co-TDMA has a precise meaning associated with TXOP sharing among coordinated APs. In IEEE 802.11bn, once an AP wins channel access, it may allocate part of its TXOP to a neighboring AP, allowing otherwise-colliding basic service sets to transmit in a conflict-free, round-robin TDMA-like manner (Lee et al., 26 Aug 2025). In IEEE 802.11be, the corresponding coordinated time division multiple access mechanism is described through a coordination group containing a Sharing AP and Shared APs; the Sharing AP gains the channel via standard DCF contention and then allocates portions of the granted TXOP to other APs in coordinated slots (Nunez et al., 2021).

Outside WLAN standardization, the term is often used more generically to denote conflict-free, globally or locally coordinated slot assignment. The literature includes centrally managed static time-slot allocation for multi-hop sensor networks, connection-oriented multi-hop mesh schedules with guaranteed time slots, and distributed reservation or state-exchange methods that converge to collision-free turn taking (Nguyen et al., 2021, Terraneo et al., 2018, Andreoli-Fang et al., 2022, Hui et al., 2011). This broader usage emphasizes the scheduling property—coordinated, interference-aware slot occupation—rather than a single protocol family.

A recurring distinction is between periodic and non-periodic coordination. Classical TDMA commonly implies a fixed periodic schedule, whereas several coordinated schemes are explicitly dynamic. Dynamic Time Slot Allocation (DTSA) for quadcopter swarms is decentralized and non-periodic; the next sender is reselected at every slot according to communication demand and collision risk, rather than by round robin (Azem et al., 2022). This distinction matters because much of the recent Co-TDMA literature addresses how coordination is updated under congestion, mobility, or heterogeneous traffic.

2. Coordination architectures

Centralized Co-TDMA architectures assign a distinguished coordinator responsibility for topology acquisition, slot computation, and schedule dissemination. In vehicular RTS-TDMA, a roadside unit (RSU) acts as the coordinator within its service zone, collects all mini-slot observations in the contention access phase, applies successive interference cancellation (SIC), extracts vehicle IDs, and announces which vehicles are permitted to transmit in which contention-free slots (Peng et al., 2019). In TDMH-MAC, a master node performs periodic clock synchronization, gathers the full network topology, computes the global TDMA schedule, and disseminates it over the mesh using Glossy flooding (Terraneo et al., 2018). In the 3-egress gateway linear-topology sensor network, a central management server knows or estimates topology and link parameters, computes static global slot allocations, and distributes the schedule to relay nodes (Nguyen et al., 2021).

Centralized control also appears in hybrid contention/scheduled systems. The MDP-based centralized channel access model (MCCA) for single-hop IEEE 802.15.4-like networks maintains buffer state and traffic information for all nodes, centrally decides which nodes receive TDMA slots in the contention-free period (CFP), which may contend in the contention access period (CAP), and which should defer transmission, then broadcasts those instructions in the beacon (Shrestha et al., 2014). These schemes exploit global state for throughput, congestion relief, or energy efficiency, but require explicit state collection and nontrivial control overhead.

Distributed Co-TDMA architectures remove the single coordinator but retain explicit slot coordination. MAC-RSV is synchronous and distributed: nodes contend in a signaling phase, reserve slots by RTS-(N)CTS-CONF exchanges, and then validate the reservation by a receiver beacon at the start of each data slot (Andreoli-Fang et al., 2022). The peer-to-peer state-exchange protocols of Leith and Clifford assume that one-hop peers periodically exchange a small amount of state information; each station then adapts its state and packet length so that stations take turns to transmit in each neighborhood without collision (Hui et al., 2011). Local multicoloring algorithms push decentralization further: each node knows only its own ID and the IDs of immediate neighbors and computes a nearly optimal TDMA schedule in constant time (0902.1868).

Hybrid architectures combine contention and coordination. In the hybrid CSMA/CA-TDMA schemes, nodes request TDMA slots during CAP and use them during CFP; the distributed model (MDCA) uses a per-node MDP based on local buffer state, while the centralized model uses the joint buffer state vector of all nodes (Shrestha et al., 2014). RTS-TDMA likewise splits request and transmission, using contention only for short request packets and reserving full safety-message transmission for contention-free slots (Peng et al., 2019). These designs preserve some of the flexibility of contention-based access while reserving deterministic service for traffic that benefits from explicit coordination.

3. Temporal structure and control signaling

A defining property of Co-TDMA systems is the explicit separation of coordination signaling from payload transmission. RTS-TDMA divides each frame into three sequential phases: a contention access phase (CAP), a broadcast feedback phase (BFP), and a contention-free transmission phase (CTP). With $N_c$ CAP mini-slots of duration $\tau_c$, $N_t$ CTP slots of duration $\tau_t$, and frame duration $T_F$, the frame relation is

$N_c\tau_c + T_f + N_t\tau_t = N_I\tau_I = T_F.$

Vehicles send only their ID and slot-selection pointers in the CAP, the RSU broadcasts the list of scheduled IDs in the BFP, and only scheduled vehicles transmit full safety messages in the CTP (Peng et al., 2019).

Reservation-based multihop designs use similar two-stage organizations. MAC-RSV frames consist of a signaling phase and a data transmission phase. The signaling phase contains $K$ mini-slot triplets for collective reservation of data slots: RTS, CTS or NCTS, and CONF. Each data slot begins with a receiver beacon (RB), followed by payload transmission and an ACK mini-slot (Andreoli-Fang et al., 2022). The RB is not merely an acknowledgment mechanism; it also detects deadlock or mobility-induced conflicts and informs newly appearing neighbors that the slot is reserved.

Centralized mesh protocols integrate synchronization and control dissemination into the time structure. TDMH-MAC divides time into tiles; each tile starts with a control slot, either downlink or uplink, followed by data slots. Synchronization frames include a 32-bit counter incremented each period, and synchronization packets are sent every 10–60 s using FLOPSYNC-2. Schedule dissemination is then performed by Glossy flooding, after which each node extracts its own schedule fragment (Terraneo et al., 2018). The resulting data superframe is deterministic and connection-oriented.

In WLAN Co-TDMA, coordination is embedded inside TXOP sharing procedures rather than superframes. In IEEE 802.11be c-TDMA, channel reservation uses MAP-RTS and MAP-CTS frames, and slot assignment uses MAP-TF trigger frames (Nunez et al., 2021). In IEEE 802.11bn, the sharing AP first attempts its own downlink or uplink low-latency traffic; if time remains and a candidate shared AP responded during polling, it allocates a shared TXOP via Co-TDMA, after initial control frame exchanges and a MU-RTS TXOP Sharing trigger (Lee et al., 26 Aug 2025). The signaling path is therefore shorter-lived than in superframe systems but conceptually analogous: reservation is explicit, and access within the reserved interval is scheduled rather than contended.

4. Scheduling objectives and analytical formulations

The analytical core of Co-TDMA varies by application, but most formulations optimize one of four quantities: latency, throughput, reliability, or bandwidth guarantees under interference constraints.

In IEEE 802.11bn, the latency benefit of Co-TDMA is expressed through the channel access delay gain: $$T^{\text{Access-delay-gain}_i} = T^{\text{Access-interval}_{u,i}} - T^{\text{Access-interval}_{c,i}}.$$ For high-congestion scenarios, this is approximated as

$$T^{\text{Access-delay-gain}_i} \approx T^{\text{Busy}_{u,i}} - T^{\text{Co-TDMA}},$$

capturing the reduction in idle waiting when a shared AP transmits inside another AP’s TXOP instead of waiting for an independent access opportunity (Lee et al., 26 Aug 2025).

Reliability-oriented Co-TDMA formulations explicitly allocate redundancy. In the 3-egress gateway linear-topology sensor network, if $s_{i,j}$ denotes the number of slots allocated to the packet of node $\tau_c$0 on link $\tau_c$1, the delivery probability of a packet from node $\tau_c$2 is

$\tau_c$3

and the joint probability that all packets from all nodes are delivered is

$\tau_c$4

A relaxed slot-allocation solution is obtained through a Lagrangian method, yielding

$\tau_c$5

with $\tau_c$6 determined by the slot-sum constraint (Nguyen et al., 2021).

Risk-aware decentralized scheduling uses kinematic state rather than queue state. In DTSA, each quadcopter computes for every active agent $\tau_c$7

$\tau_c$8

where

$\tau_c$9

The next sender is the agent with the largest priority value, with ties broken by a counter tracking time since last transmission (Azem et al., 2022). The objective is not static fairness but rapid dissemination of state for agents with the highest collision risk or maneuver urgency.

Queue-aware coordinated access can be posed as an MDP. In the hybrid CSMA/CA-TDMA schemes, a node or coordinator chooses among four actions—defer transmission, CAP transmission, CFP transmission, or both CAP and CFP—based on buffer occupancy. The optimal policy satisfies

$N_t$0

and the centralized state-action space grows as $N_t$1 (Shrestha et al., 2014). This formulation makes explicit the trade-off between throughput, energy, and bandwidth efficiency.

Graph-theoretic distributed formulations make the schedule itself the object of optimization. In one-shot local multicoloring, each node $N_t$2 of degree $N_t$3 receives a set $N_t$4 of colors such that neighboring nodes receive disjoint sets, with guarantee

$N_t$5

This is a direct local approximation to a TDMA slot-set assignment under interference constraints (0902.1868). In edge computing offloading, sequence optimization gives a different but related perspective: the TDMA slot order is a decision variable $N_t$6, and for identical sensing rates the optimal sequence follows the ascending order of weighted transmission rates $N_t$7 (Liang et al., 2021). A plausible implication is that Co-TDMA design increasingly treats slot order itself as a schedulable resource, not merely slot ownership.

5. Representative empirical behavior

Empirical results show that coordinated time scheduling is not associated with a single performance signature; its effect depends strongly on whether the coordination mechanism is paired with traffic awareness, redundancy, spatial reuse, or request/data separation.

For IEEE 802.11bn real-time traffic, Co-TDMA lowers the 95th percentile end-to-end latency and the standard deviation of end-to-end latency. In the reported RTMG scenarios, worst-case latency improves by approximately 12–24% depending on network congestion, and the paper summarizes an improvement of approximately 24% for worst-case latency while also reporting lower jitter and negligible impact on overall network throughput (Lee et al., 26 Aug 2025). In IEEE 802.11be, however, plain c-TDMA is reported to be similar to or slightly below non-coordinated multi-AP throughput, with maximum gain close to 5%; the substantial gains arise only when c-TDMA is combined with spatial reuse, where throughput gain exceeds 140% in 90% of considered scenarios for the 4-AP case (Nunez et al., 2021).

In dense vehicular networks, request-transmission splitting changes the trade-off between control overhead and payload collisions. RTS-TDMA uses short request packets, repetition according to a distribution such as $N_t$8, and SIC at the RSU; both theoretical analysis and numerical results verify higher throughput than coded slotted ALOHA. Under the paper’s practical values with $N_t$9 ms, $\tau_t$0 ms, and $\tau_t$1 ms, RTS-TDMA clearly outperforms CSA when the service-zone size satisfies $\tau_t$2 (Peng et al., 2019).

In aerial swarms, dynamic decentralized coordination changes the scalability regime. DTSA works for up to 70 quadcopters when all are moving, whereas periodic TDMA fails for $\tau_t$3 in the same setting. When $\tau_t$4, DTSA provides up to 5–9 times more frequent updates for active flyers than TDMA, and the reported simulations and Crazyflie experiments show zero collisions up to the tested maximum $\tau_t$5, larger minimum inter-agent distance, smoother trajectories, and shorter completion time than TDMA (Azem et al., 2022).

In multi-hop deterministic networks, centralized coordination is associated with high utilization and explicit reliability control. TDMH-MAC reports well over 80–90% data-slot usage with 100 ms tiles in simulations up to 32 nodes and 5–6 hop networks, and in a 9-node deployment stream reliability rises from 95.4–99.6% without redundancy to 99.0–99.97% with spatial redundancy (Terraneo et al., 2018). MAC-RSV, though distributed rather than centralized, shows how reservation-based Co-TDMA can dominate contention-based access in static and mobile settings: aggregate throughput approaches 18 Mbps in the 25-node static mesh, compared with about 4 Mbps for IEEE 802.11, is 70% higher than 802.11 in the mobile case, and is over 55% higher than CATA under contention (Andreoli-Fang et al., 2022).

6. Limitations, distinctions, and research directions

A common misconception is that Co-TDMA is synonymous with static round-robin slotting. The literature does not support that reduction. DTSA is explicitly non-periodic and demand-adaptive (Azem et al., 2022); RTS-TDMA dynamically resolves requests each frame (Peng et al., 2019); hybrid CSMA/CA-TDMA schemes switch nodes among CAP, CFP, both, or sleep according to MDP policies (Shrestha et al., 2014). Even in Wi-Fi TXOP sharing, the decision to allocate a shared TXOP depends on residual TXOP duration and the presence of a responding candidate shared AP (Lee et al., 26 Aug 2025).

A second misconception is that collision-free coordination necessarily improves throughput. IEEE 802.11be c-TDMA alone does not result in significant throughput improvement because coordination overheads and sequential slotting offset the gain from eliminating inter-AP contention; the large gains appear only when the same coordination substrate enables coordinated spatial reuse (Nunez et al., 2021). The broader record therefore distinguishes between coordination as an access-control primitive and coordination as a complete performance solution.

Centralization and distribution impose different limits. Centralized models such as MCCA outperform distributed ones but require global information and face exponential state-action growth (Shrestha et al., 2014). TDMH-MAC depends on sub-microsecond synchronization, topology reporting, and schedule dissemination, with topology convergence times below 100 s for practical deployments up to 32 nodes and below 5 min for 128 nodes (Terraneo et al., 2018). Distributed TDMA-like protocols are lighter-weight but rely on assumptions about local state exchange and topology stability; the peer-to-peer state-exchange protocols converge to collision-free turn taking as long as the topology remains static or changes slowly with respect to protocol execution (Hui et al., 2011), while one-shot multicoloring assumes unique IDs and one-hop neighborhood knowledge (0902.1868).

The main contemporary direction is therefore not simply “more coordination,” but richer coordination. The empirical and analytical record associates the strongest gains with control-data separation, redundancy, queue- or risk-aware slot assignment, or spatial reuse layered on top of a coordinated time structure (Peng et al., 2019, Nguyen et al., 2021, Azem et al., 2022, Nunez et al., 2021). This suggests that Co-TDMA is best understood as a scheduling substrate whose practical value depends on how effectively the surrounding system exploits topology knowledge, traffic asymmetry, interference structure, and real-time constraints.