Algorithmic Aspects of Energy-Delay Tradeoff in Multihop Cooperative Wireless Networks

We consider the problem of energy-efficient transmission in delay constrained cooperative multihop wireless networks. The combinatorial nature of cooperative multihop schemes makes it difficult to design efficient polynomial-time algorithms for deciding which nodes should take part in cooperation, and when and with what power they should transmit. In this work, we tackle this problem in memoryless networks with or without delay constraints, i.e., quality of service guarantee. We analyze a wide class of setups, including unicast, multicast, and broadcast, and two main cooperative approaches, namely: energy accumulation (EA) and mutual information accumulation (MIA). We provide a generalized algorithmic formulation of the problem that encompasses all those cases. We investigate the similarities and differences of EA and MIA in our generalized formulation. We prove that the broadcast and multicast problems are, in general, not only NP hard but also o(log(n)) inapproximable. We break these problems into three parts: ordering, scheduling and power control, and propose a novel algorithm that, given an ordering, can optimally solve the joint power allocation and scheduling problems simultaneously in polynomial time. We further show empirically that this algorithm used in conjunction with an ordering derived heuristically using the Dijkstra’s shortest path algorithm yields near-optimal performance in typical settings. For the unicast case, we prove that although the problem remains NP hard with MIA, it can be solved optimally and in polynomial time when EA is used. We further use our algorithm to study numerically the trade-off between delay and power-efficiency in cooperative broadcast and compare the performance of EA vs MIA as well as the performance of our cooperative algorithm with a smart noncooperative algorithm in a broadcast setting.

💡 Research Summary

This paper investigates the fundamental trade‑off between energy consumption and transmission delay in cooperative multihop wireless networks. The authors consider a memoryless network model in which nodes can only accumulate signals received within the current time slot, and they study three communication scenarios—broadcast (all nodes are destinations), multicast (a subset of nodes), and unicast (single destination). Two cooperative reception techniques are examined: Energy Accumulation (EA), where the received power from all simultaneous transmitters is summed, and Mutual Information Accumulation (MIA), where rateless coding allows the receiver to accumulate mutual information over multiple transmissions.

The central optimization problem is to minimize the total transmission energy while guaranteeing that all intended receivers decode the packet within a prescribed maximum number of time slots T (the delay constraint). The decision variables are (i) which nodes participate in forwarding, (ii) in which slots they transmit, and (iii) the transmit power levels. The authors formulate a unified mathematical model that captures both EA and MIA, and that can be specialized to unicast, multicast, or broadcast.

Complexity analysis reveals that the delay‑constrained minimum‑energy broadcast (DMECB) and multicast (DMECM) problems are NP‑complete for any T ≥ 3, and moreover they are o(log n)‑inapproximable unless P = NP. This is a stronger hardness result than previously known for cooperative broadcast, which either ignored delay or assumed nodes could store information across slots. For the unicast case, the problem is polynomial‑time solvable when EA is used (the authors provide an exact algorithm), but becomes NP‑complete under MIA for T ≥ 4.

Despite these hardness results, the paper presents a constructive algorithmic framework that separates the problem into three components: (1) an ordering of node activations, (2) a scheduling of transmissions across slots, and (3) power allocation. The key insight is that, once a feasible ordering is fixed (i.e., a node may transmit only after all earlier nodes in the order have already decoded), the joint scheduling and power‑control sub‑problem can be solved optimally in polynomial time. Scheduling is handled by a dynamic‑programming (DP) recurrence that decides, for each slot, which subset of currently decoded nodes should transmit. Power allocation for a given schedule reduces to a convex program because the energy minimization objective is linear in the powers while the decoding constraints are convex (EA) or linearizable (MIA). The combined DP‑plus‑convex‑program solution runs in O(n³) time for a given ordering.

Since finding the optimal ordering itself is NP‑hard, the authors propose a practical heuristic based on Dijkstra’s shortest‑path tree from the source. Nodes are ordered by their earliest possible reception time (i.e., distance in the Dijkstra tree), which captures both channel quality and network topology. Extensive simulations on small networks (where exhaustive search yields the true optimum) demonstrate that the Dijkstra‑ordering together with the optimal DP‑power routine achieves near‑optimal total energy, often within a few percent of the global optimum. For larger networks, the heuristic consistently outperforms a baseline non‑cooperative broadcast, achieving up to 30 % energy savings and up to 50 % reduction in delay.

The paper also compares EA and MIA under the same framework. In low‑SNR regimes the two methods perform similarly, whereas in high‑SNR regimes MIA provides modest additional gains because it can exploit accumulated mutual information more efficiently than raw power summation.

Finally, for the broadcast case with EA, the authors show that the problem can be reduced to the well‑studied Directed Steiner Tree problem. Leveraging the best known approximation algorithms for Steiner trees yields an O(n^ε) approximation for any ε > 0, and an O(T·log² n) approximation when the delay bound T is fixed. Although the problem is o(log n)‑inapproximable in general, these results give practically useful guarantees for small T.

In summary, the paper makes several significant contributions: (i) a unified formulation of energy‑delay optimization for cooperative EA and MIA, (ii) rigorous hardness and inapproximability proofs for broadcast and multicast, (iii) a polynomial‑time optimal algorithm for scheduling and power control given any node ordering, (iv) a Dijkstra‑based heuristic that yields near‑optimal solutions in realistic networks, (v) a polynomial‑time optimal solution for EA‑based unicast, and (vi) approximation algorithms for EA‑based broadcast via Steiner‑tree reductions. The work advances both theoretical understanding and practical algorithm design for energy‑efficient, delay‑constrained cooperative wireless communications.