Markovian Battery State Modeling for Energy Harvesting Communications

21 concepts17 papers1 proposal

In traditional wireless communications, we assume devices have a constant, reliable power supply. If a transmitter needs to send a packet, it draws power from an effectively bottomless grid or a large battery, and the energy cost per bit remains a static engineering constraint.

Energy harvesting (EH) changes this paradigm. When nodes scavenge energy from ambient environments—such as solar, wind, or radio-frequency (RF) signals—power availability becomes stochastic. More importantly, storing this harvested energy in a finite battery introduces a profound challenge: temporal dependency, or memory.

Whether a device can transmit a packet right now depends on its current battery level. Its current battery level depends on whether it transmitted a packet in the previous slot, which depends on the battery level before that, stretching all the way back to the system's initialization. In information theory and network analysis, this means the communication channel effectively possesses infinite memory, breaking standard memoryless models.

To make these systems tractable, researchers use Markovian battery state modeling. By treating the discrete energy levels of a finite battery as states in a Markov chain or a stochastic queue, we can capture this infinite historical dependency within a clean, state-based mathematical framework.


Mapping Energy Flow to Discrete Queues

The core intuition behind this approach is to treat energy exactly like data packets. Just as a data buffer stores incoming packets and releases them upon transmission, a battery stores incoming "energy packets" (joules or quanta of energy) and depletes them to transmit data.

In Transmission Capacity of Wireless Ad Hoc Networks with Energy Harvesting Nodes, Rahul Vaze demonstrated that energy storage can be modeled as a discrete birth-death Markov process. In this setup:

  • Arrivals (Births): Environmental energy harvesting events are modeled as a stochastic process (e.g., a Bernoulli or Poisson distribution).
  • Departures (Deaths): Data transmissions consume a discrete number of energy units from the buffer.
   Harvest Energy (+1)         Harvest Energy (+1)
       ┌─────────┐                 ┌─────────┐
       │         │                 │         │
       ▼         │                 ▼         │
   ┌───────┐     │             ┌───────┐     │             ┌───────┐
   │State 0│     │             │State 1│     │             │State 2│  ... (up to Capacity K)
   └───────┘     │             └───────┘     │             └───────┘
       │         │                 ▲         │                 ▲
       │         └─────────────────┘         └─────────────────┘
       │       Transmit Packet (-1)        Transmit Packet (-1)
       └───────────────────────────────────────────────────────┘
                         Outage (No Energy)

This abstraction allows researchers to apply queuing theory directly to physical energy storage. For example, Wang et al. modeled the energy flow of a solar-powered cellular base station as a stochastic $M/G/1/K$ queue, where $K$ represents the finite battery capacity. By framing the battery as a finite buffer where solar energy arrives randomly and traffic-driven consumption acts as the service process, the authors derived closed-form expressions for system reliability and service outage probabilities. Similarly, El Shafie et al. utilized a discrete-time queueing system to track the energy states of transmitters and friendly jammers, optimizing transmission timing to guarantee physical-layer security under fluctuating power constraints.


The Markovian Framework: Solving for Steady State

Why does framing the battery as a Discrete-Time Markov Chain (DTMC) work so well? Because it allows us to compute the steady-state probability distribution of a device's energy.

If we know the probability matrix governing how energy enters and leaves the battery, we can calculate the long-term fraction of time the device spends in any given energy state. Crucially, this tells us the probability of energy outage—the likelihood that the device finds itself in "State 0" when it needs to transmit.

In the context of cognitive radio networks, Rui Zhang 0006 and Kaibin Huang used this framework to model secondary users who scavenge RF energy from a primary network's transmissions. By modeling the secondary transmitter's battery as a finite-state Markov chain, the authors determined its steady-state probability of having sufficient stored energy to initiate a transmission without interfering with the primary network. In Opportunistic Wireless Energy Harvesting in Cognitive Radio Networks, the same authors extended this to spatial networks, using the DTMC steady-state distribution to calculate the exact density of active, self-sustaining nodes that a network can support.

When multiple devices interact, their battery states become coupled. To address this, Heydar Abad et al. modeled the joint battery states of multiple devices using a multi-dimensional DTMC. This allowed them to account for environmental correlations (e.g., two nearby sensors harvesting solar energy from the same cloud-covered sky) and derive optimal, threshold-based transmission policies based on the collective energy state of the network.


Co-Evolution of Energy and Information

In real-world Internet of Things (IoT) and sensor networks, tracking energy alone is insufficient. A node might have plenty of energy but no data to send, or it might have fresh data but an empty battery. To capture the true performance of a network, models must track the co-evolution of both energy and data.

To solve this, Gharbieh et al. introduced a two-dimensional discrete-time Markov chain. One dimension tracks the discrete energy levels in the battery, while the second dimension tracks the data packets waiting in the transmission queue. This two-dimensional state space explicitly captures the gating mechanism inherent to EH communications: transmission is authorized if and only if the data queue is non-empty and the energy queue is above the required transmission threshold.

                Data Packet Queue State ──►
             ┌───────┬───────┬───────┬───────┐
             │ (0,0) │ (0,1) │ (0,2) │ (0,3) │
             ├───────┼───────┼───────┼───────┤
  Energy     │ (1,0) │ (1,1) │ (1,2) │ (1,3) │
  Battery    ├───────┼───────┼───────┼───────┤
  State      │ (2,0) │ (2,1) │ (2,2) │ (2,3) │
    │        ├───────┼───────┼───────┼───────┤
    ▼        │ (3,0) │ (3,1) │ (3,2) │ (3,3) │
             └───────┴───────┴───────┴───────┘
              ▲
              └─ Active transmission requires both 
                 Energy > 0 AND Data > 0

This joint modeling approach is particularly vital for analyzing modern network metrics like Age of Information (AoI), which measures data freshness at the receiver.

  • Durisi et al. utilized a DTMC to analyze Slotted ALOHA networks, showing how data packet generation and stochastic energy arrivals jointly dictate the optimal transmission probability needed to minimize AoI.
  • Zhao et al. advanced this by using bulk-service Markov chains. In their model, a single transmission event could consume multiple discrete units of harvested energy depending on the channel conditions, revealing a stark trade-off: in energy-starved regimes, the freshest information is achieved by transmitting immediately when the minimum energy threshold is met, rather than waiting to pool energy.
  • Chao Zhang et al. applied finite Markov chains to cooperative relay networks, demonstrating how cluster heads can optimize their relaying behavior based on the probabilistic evolution of their batteries over time.

Finding the Limits of Channels with Infinite Memory

Beyond network protocols, Markovian battery modeling addresses a fundamental information-theoretic question: What is the absolute capacity of an energy-harvesting channel?

As established by Ozgur et al., a finite battery introduces temporal dependencies that make calculating channel capacity notoriously difficult. Because the current state is tied to the entire history of the system, standard memoryless capacity formulas (like Shannon's $C = B \log(1 + SNR)$) do not apply.

To bypass this infinite memory trap, Abdollahi Bafghi et al. framed the finite battery as a channel state with memory. By treating the battery level as an explicit state variable within a Finite State Channel (FSC) framework, the system's history is neatly encapsulated by the current state.

Using this state-space reduction, Mao et al. applied ergodic theory to derive computable upper and lower bounds for discrete energy harvesting channels. Crucially, the authors demonstrated that near-optimal transmission strategies require knowledge of only the current battery state rather than the entire history of energy arrivals, proving that low-complexity, memoryless decision-making can achieve near-optimal data rates.

This mathematical tractability has led to several surprising insights into how energy constraints impact communication limits:

  1. The Power of Feedback: In classical memoryless channels, feedback from the receiver does not increase capacity. However, Shaviv et al. proved that for energy-harvesting channels, feedback does increase capacity. Because the channel has memory via the battery, feedback allows the transmitter to infer the receiver's state and better manage its remaining energy reserves.
  2. The "Good Enough" Battery Threshold: Designers often worry about optimizing battery size. However, Shaviv et al. developed a universal approximation formula showing that a battery capable of storing just the maximum amount of energy arrived in a single time slot is sufficient to achieve near-optimal capacity.

By transforming the physical reality of a fluctuating, finite battery into a structured Markovian state space, this research direction bridges the gap between unpredictable environmental physics and rigorous communication theory.

Key Foundational Work

Cognitive Energy Harvesting and Transmission from a Network Perspective (2012) — contributed 1 concept elements to this direction.
Transmission Capacity of Wireless Ad Hoc Networks with Energy Harvesting Nodes (2012) — contributed 1 concept elements to this direction.
Analytical models for energy consumption in infrastructure WLAN STAs carrying TCP traffic (2010) — contributed 1 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_markovian_battery_state_modeling_b704b6c1,
  title = {Markovian battery state modeling},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/markovian-battery-state-modeling/rdr_b704b6c15c3b41fcace76802d94b21fa},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-20},
}

This exact version (version 1, frozen text):

@misc{lacuna_markovian_battery_state_modeling_b704b6c1_v1,
  title = {Markovian battery state modeling},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/markovian-battery-state-modeling/rdr_b704b6c15c3b41fcace76802d94b21fa/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-20},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake