[Paper Review] Throughput Maximization for an Energy Harvesting Communication System with Processing Cost

Oner Orhan, Denız Gündüz+1 more|arXiv (Cornell University)|Aug 23, 2012

Energy Harvesting in Wireless Networks9 references7 citations

TL;DR

This paper proposes an optimal offline transmission policy for an energy-harvesting transmitter over a fading channel, jointly accounting for transmission and processing energy costs. Using a directional backward glue-pouring algorithm, it maximizes throughput under energy causality and finite battery constraints, showing that processing costs induce bursty transmission and reduce total throughput as processing energy increases.

ABSTRACT

In wireless networks, energy consumed for communication includes both the transmission and the processing energy. In this paper, point-to-point communication over a fading channel with an energy harvesting transmitter is studied considering jointly the energy costs of transmission and processing. Under the assumption of known energy arrival and fading profiles, optimal transmission policy for throughput maximization is investigated. Assuming that the transmitter has sufficient amount of data in its buffer at the beginning of the transmission period, the average throughput by a given deadline is maximized. Furthermore, a "directional glue pouring algorithm" that computes the optimal transmission policy is described.

Motivation & Objective

• Address the gap in energy-harvesting communication systems by jointly modeling transmission and processing energy costs, which are often ignored in prior work.
• Formulate an offline optimization problem to maximize total transmitted data by a deadline under energy causality and finite battery capacity.
• Identify the optimal transmission policy that balances energy usage across time slots, considering both channel fading and processing energy costs.
• Develop a computationally efficient algorithm to compute the optimal power allocation and transmission duration for each time slot.
• Analyze the impact of processing energy cost on system throughput and transmission structure, showing a shift from continuous to bursty transmission.

Proposed method

• Formulate the throughput maximization problem as a convex optimization problem with constraints on energy causality, finite battery capacity, and non-negative power and duration.
• Introduce a dual variable approach to derive optimality conditions, where the optimal power level in each epoch is determined by a threshold related to the inverse channel gain and processing cost.
• Propose a 'directional backward glue-pouring' algorithm that allocates harvested energy starting from the last non-zero energy arrival backward in time, ensuring energy causality and optimal power allocation.
• Use the glue-pouring principle to allocate energy across epochs such that the optimal power level is maintained in partially used epochs, while fully used epochs operate above the threshold.
• Incorporate processing energy cost ε as a constant per-unit-time cost, which affects the effective power threshold and leads to bursty transmission when ε is non-zero.
• Solve the optimization problem using KKT conditions, ensuring a unique solution when constraints are active or dual variables are non-zero.

Experimental results

Research questions

• RQ1 How does the inclusion of processing energy cost affect the optimal transmission policy in an energy-harvesting system with fading channels?
• RQ2 What is the structure of the optimal transmission policy when both transmission and processing energy costs are considered?
• RQ3 How does the finite battery capacity interact with processing energy costs to shape the optimal power allocation and transmission duration?
• RQ4 What is the performance gain of using noncausal knowledge of energy arrivals and channel states in maximizing throughput under processing energy costs?
• RQ5 How does increasing the processing energy cost ε affect the total achievable throughput and the burstiness of the transmission schedule?

Key findings

• The optimal transmission policy becomes bursty when processing energy cost ε > 0, as opposed to continuous transmission when ε = 0, due to the trade-off between processing and transmission energy.
• For ε = 0, the optimal policy fully utilizes each epoch with power levels determined by energy causality and channel gains, as shown in Fig. 2a with total throughput of 2.11 nats.
• For ε = 1 μW, the optimal policy reduces total transmission duration and energy usage, resulting in a lower total throughput of 1.39 nats, as shown in Fig. 2b.
• The directional backward glue-pouring algorithm efficiently computes the optimal policy by allocating energy from the last non-zero energy arrival backward, respecting battery capacity and causality.
• The processing energy cost causes energy to be allocated to worse channel states (e.g., epoch 2 with h=0.2×10⁶) when battery capacity is limited, to avoid high processing costs from prolonged transmission.
• Throughput decreases monotonically with increasing processing energy cost ε, as shown in Fig. 3, confirming the trade-off between processing and transmission energy efficiency.