A Distributed and Deterministic TDMA Algorithm for Write-All-With-Collision Model

Several self-stabilizing time division multiple access (TDMA) algorithms are proposed for sensor networks. In addition to providing a collision-free communication service, such algorithms enable the transformation of programs written in abstract models considered in distributed computing literature into a model consistent with sensor networks, i.e., write all with collision (WAC) model. Existing TDMA slot assignment algorithms have one or more of the following properties: (i) compute slots using a randomized algorithm, (ii) assume that the topology is known upfront, and/or (iii) assign slots sequentially. If these algorithms are used to transform abstract programs into programs in WAC model then the transformed programs are probabilistically correct, do not allow the addition of new nodes, and/or converge in a sequential fashion. In this paper, we propose a self-stabilizing deterministic TDMA algorithm where a sensor is aware of only its neighbors. We show that the slots are assigned to the sensors in a concurrent fashion and starting from arbitrary initial states, the algorithm converges to states where collision-free communication among the sensors is restored. Moreover, this algorithm facilitates the transformation of abstract programs into programs in WAC model that are deterministically correct.

💡 Research Summary

The paper addresses the problem of assigning collision‑free communication slots in sensor networks that operate under the Write‑All‑With‑Collision (WAC) model. In the WAC model a sensor’s write operation updates the state of all its neighbors atomically, but if multiple neighbors write simultaneously the target’s state remains unchanged, effectively modelling broadcast collisions. Existing self‑stabilizing TDMA (time‑division multiple access) algorithms either rely on randomization, assume prior knowledge of the network topology, or recover slots sequentially through a distinguished node (e.g., a base station). These limitations hinder deterministic program transformation from abstract distributed models to concrete WAC implementations, restrict dynamic node addition, and cause slow convergence.

The authors propose a deterministic, fully distributed, self‑stabilizing TDMA slot assignment algorithm that requires each sensor to know only its immediate neighbors. The core idea is to treat TDMA slot allocation as a distance‑2 graph coloring problem: each sensor’s color corresponds to its initial slot, and the TDMA period is set to the number of colors (ideally d² + 1, where d is the maximum node degree). When a sensor detects a persistent collision on a particular slot, it initiates a “reset” computation that locally recolors the sensors within its distance‑2 neighborhood, thereby eliminating the conflict.

The reset mechanism is carefully timed to avoid interference between concurrent resets. A sensor j that observes a collision records the colliding slot and the frame number, then computes a unique reset frame f_reset = f_collision + ID_j + D₃_timeout. The addition of its identifier guarantees that two sensors detecting collisions in the same frame schedule resets in different frames. Before transmitting the reset message, j remains silent for D₃_timeout frames, ensuring that its distance‑3 neighborhood has also stopped transmitting; this prevents the reset and subsequent color‑change messages from colliding with ordinary traffic.

The reset proceeds in five steps:

1. Collision detection & reset scheduling – sensors maintain two tables, nbrClr (1‑hop neighbor colors) and dist2Clr (2‑hop neighbor colors), and schedule a reset when a threshold of consecutive collisions is reached.
2. Reset message broadcast – the initiator sends a reset packet in its allocated reset slot; neighbors update their tables upon receipt.
3. Color update – each neighbor checks whether its current color conflicts with any distance‑2 neighbor. If so, it selects the smallest unused color (slot) from the palette, guaranteeing a conflict‑free assignment within the local 2‑hop region.
4. Change‑color notification – the node that changed its color broadcasts a ChangeColor message in its newly assigned slot. All affected nodes update their tables and cancel any pending resets that would conflict with the new assignment.
5. Restart communication – the initiator finally sends a Restart message, signalling the distance‑3 neighborhood that normal TDMA operation may resume.

Because each reset eliminates at least one collision, repeated resets converge to a legitimate state where no two nodes within distance 2 share a slot. The algorithm’s self‑stabilizing property guarantees convergence from any arbitrary initial coloring, even after arbitrary transient faults. The worst‑case convergence time is bounded by the product of the network diameter and the maximum number of colors, but the design prevents exponential message explosion: only one reset can be active in any distance‑2 neighborhood at a time, and the D₃_timeout ensures silence in the surrounding area during the critical phases.

Dynamic network changes are supported without redesign. When a node leaves, its slot becomes free and neighboring nodes can reclaim it during subsequent resets. When a new node joins, it learns the colors of its 1‑hop and 2‑hop neighbors (through existing neighbor broadcasts) and selects the smallest unused color, triggering a local reset only if necessary. The authors also discuss extensions for non‑uniform bandwidth allocation (assigning multiple slots to high‑traffic nodes) and for clock synchronization, noting that the algorithm can be combined with existing time‑synchronization protocols to maintain slot alignment.

The paper concludes with a comparative analysis showing that, unlike prior deterministic TDMA schemes that require a central coordinator or sequential recovery, the proposed method achieves concurrent, deterministic recovery while preserving the same color bound (d² + 1). This makes it suitable for transforming abstract distributed algorithms into concrete WAC‑model implementations with guaranteed correctness, deterministic behavior, and support for dynamic topology changes—key requirements for real‑world sensor network deployments.