Energy-Autonomous Wireless Sensing Node (EAWSN)

• EAWSN is a self-powered wireless sensor node that integrates energy harvesting, storage, power management, and adaptive communication to support battery-less operation.
• It employs an energy-driven architecture to balance individual, local, global, and environmental energy budgets under energy-neutrality conditions.
• Designs focus on co-optimizing sensing and communication processes and addressing trade-offs in storage, protocol efficiency, and hardware-specific limitations.

An Energy-Autonomous Wireless Sensing Node (EAWSN) is a wireless sensing node designed to sustain sensing, processing, storage, and communication from harvested energy rather than periodic battery replacement. In the literature, autonomy is usually expressed through an energy-neutrality condition in which long-horizon harvested power covers long-horizon consumption, typically written as $$\overline{P_{\text{harvest}}} \ge \overline{P_{\text{consume}}}$$, while practical realizations combine an energy harvester, storage, power management, sensing circuitry, and an energy-aware communication stack (Hoang et al., 2012, Stricker et al., 30 Jun 2025, Martins et al., 2017, Raza et al., 2016). The term encompasses battery-assisted and battery-free nodes, indoor-light and RF-powered nodes, piezoelectric harvesters, and protocol-level architectures that adapt sensing and networking behavior to volatile energy availability (Landivar et al., 2024, Costanza et al., 22 Jul 2025, Kassan et al., 2019, Setiawan et al., 2017).

1. Architectural scope and constituent decomposition

A foundational systems view is given by the Energy Driven Architecture (EDA), which decomposes wireless sensor network energy into five constituents: Individual, Local, Global, Sink, and Environment. The Individual constituent covers the processing unit, sensing unit, memory unit, and transceiver/radio unit; the Local constituent covers neighbor monitoring, local security, idle listening, collision-driven retransmissions, overhearing, and local protocol overhead; the Global constituent covers topology maintenance, routing, global control overhead, and packet-loss costs; the Sink constituent covers sink communication and sink-issued directives; and the Environment constituent models harvested energy as a positive term that offsets battery drain (Hoang et al., 2012). This decomposition is explicitly intended to replace layer-centric optimization, because minimizing one constituent in isolation can increase energy in another.

Concrete EAWSN implementations instantiate this abstraction with markedly different hardware organizations. RF-powered nodes combine a rectenna, supercapacitor storage, and a sensor mote on a common bus without a DC-DC converter to avoid conversion overhead (Setiawan et al., 2017). RF-harvesting circuit-centric designs further separate the EAWSN into an antenna subsystem, an RF Energy Harvester with an orthogonally switching charge-pump rectifier, a DC-DC converter with Maximum Power Point Tracking and cold-start, storage capacitors, a co-designed receiver front-end, and a low-power transmitter (Martins et al., 2017). Other architectures duplicate both communication and storage resources: a short-range BLE transceiver can coexist with a secondary nRF24L01 radio, while a Hybrid Energy Storage System combines a battery and a supercapacitor so that the supercapacitor absorbs TX/RX bursts and the battery supplies the baseline load (Borza et al., 2019).

At the opposite end of the integration spectrum, battery-free indoor-light nodes collapse sensing and energy storage into a single capacitor-centric design. In the resistive-sensor EAWSN using Time-Domain to Digital Conversion, the photovoltaic source charges a $C_{\text{stor}} = 440\,\mu\text{F}$ capacitor, and the same storage node is deliberately discharged through the sensor during measurement (Costanza et al., 22 Jul 2025). This architecture differs from conventional sensor-front-end separation, but it remains consistent with the broader EAWSN objective: every subsystem is organized around the energy budget of the node rather than around a fixed protocol stack.

2. Energy accounting and autonomy criteria

Formal energy accounting in EAWSN research is usually state-based or cycle-based. EDA adopts interval accounting,

$E(\Delta t) = C_E \,\Delta t,$

and refines it for the Individual constituent as

$$E_{\text{individual}, i}(\Delta t) = \sum_{u \in U} \sum_{s \in S} \Big(e_{u,s}\, t_{u,s}\Big) + \sum_{u \in U} \sum_{w \in W} e_{u,w},$$

with $U = \{Pu, Su, Mu, TRu\}$, $$S = \{\text{sleep}, \text{awake}, \text{active}, \text{idle}\}$$, and transition set $$W = \{\text{is}, \text{sw}, \text{wa}, \text{ai}, \text{ia}\}$$ (Hoang et al., 2012). The emphasis on idle leakage and switching cost is central: duty cycling is not free, and repeated transitions can offset the gains of aggressive sleeping.

A complementary formulation appears in harvest-store-use nodes. E-WAN models the stored energy of a supercapacitor-based node as

$$E_{\text{cap}}(t+1)=\max(\min(E_{\text{cap}}(t)+E_{\text{harv}}(t)-E_{\text{used}}(t),B),0),$$

with usable storage capacity $B = 0.7\,\text{J}$ on the evaluated DPP3e platform, and estimates stored energy from capacitor voltage using

$E_{\text{cap}}(t)=\frac{1}{2}CV_{\text{cap}}(t)^2$

(Stricker et al., 30 Jun 2025). Capacitor-energy relations of the same form recur in battery-free sensing nodes and RF-powered nodes, where $C_{\text{stor}} = 440\,\mu\text{F}$0 is the operative quantity for both storage sizing and phase scheduling (Costanza et al., 22 Jul 2025, Setiawan et al., 2017).

Across the literature, energy autonomy is consistently framed as an average-power or horizon-based balance. The RF-harvester circuit work states the autonomy condition as $C_{\text{stor}} = 440\,\mu\text{F}$1 (Martins et al., 2017). The wake-up-receiver and model-based-sensing work writes the same requirement over a horizon $C_{\text{stor}} = 440\,\mu\text{F}$2 with harvesting and storage efficiencies,

$C_{\text{stor}} = 440\,\mu\text{F}$3

(Raza et al., 2016). In EDA, the equivalent neutrality condition is not given as a formal theorem, but the architecture explicitly implies that harvested energy must offset the sum of Individual, Local, Global, and Sink expenditures over the planning horizon (Hoang et al., 2012). This suggests that “autonomy” is not merely the presence of an energy harvester; it is the sustained satisfaction of a budget that includes protocol overhead, state transitions, and storage dynamics.

3. Harvesting sources, storage, and power-management mechanisms

EAWSN implementations span multiple harvesting modalities. EDA explicitly lists solar, thermal, kinetic, and vibration harvesting as the Environment constituent (Hoang et al., 2012). Piezoelectric harvesting is modeled as a one-degree-of-freedom mass–spring–damper system,

$C_{\text{stor}} = 440\,\mu\text{F}$4

with control wrapped around the harvester by a PID law

$C_{\text{stor}} = 440\,\mu\text{F}$5

to regulate residual energy or storage voltage (Kassan et al., 2019). In the reported simulations, Ziegler–Nichols tuning produced $C_{\text{stor}} = 440\,\mu\text{F}$6, $C_{\text{stor}} = 440\,\mu\text{F}$7, and $C_{\text{stor}} = 440\,\mu\text{F}$8, improving the step response from rise time $C_{\text{stor}} = 440\,\mu\text{F}$9 and settling time $E(\Delta t) = C_E \,\Delta t,$0 to rise time $E(\Delta t) = C_E \,\Delta t,$1 and settling time $E(\Delta t) = C_E \,\Delta t,$2 (Kassan et al., 2019).

Indoor-light harvesting dominates several recent batteryless realizations. One BLE/LIoT platform uses an Epishine LEH3 energy-harvesting unit with an e-peas AEM10941 PMIC and a $E(\Delta t) = C_E \,\Delta t,$3 CAP-XX GA230F supercapacitor (Landivar et al., 2024). Another battery-free BLE resistive-sensor node uses a Panasonic AM-1606C amorphous-silicon photovoltaic cell and two capacitors totaling $E(\Delta t) = C_E \,\Delta t,$4; the microcontroller stop-mode quiescent current is approximately $E(\Delta t) = C_E \,\Delta t,$5, corresponding to $E(\Delta t) = C_E \,\Delta t,$6 (Costanza et al., 22 Jul 2025). At $E(\Delta t) = C_E \,\Delta t,$7 lux, the AM-1606C provides approximately $E(\Delta t) = C_E \,\Delta t,$8, while at $E(\Delta t) = C_E \,\Delta t,$9 lux the reported harvested power is approximately $$E_{\text{individual}, i}(\Delta t) = \sum_{u \in U} \sum_{s \in S} \Big(e_{u,s}\, t_{u,s}\Big) + \sum_{u \in U} \sum_{w \in W} e_{u,w},$$0 (Costanza et al., 22 Jul 2025).

RF-powered and RF-harvesting EAWSNs emphasize conversion efficiency at very low input powers. The RF energy harvester and DC-DC chain described in the circuit-centric work operates from $$E_{\text{individual}, i}(\Delta t) = \sum_{u \in U} \sum_{s \in S} \Big(e_{u,s}\, t_{u,s}\Big) + \sum_{u \in U} \sum_{w \in W} e_{u,w},$$1 to $$E_{\text{individual}, i}(\Delta t) = \sum_{u \in U} \sum_{s \in S} \Big(e_{u,s}\, t_{u,s}\Big) + \sum_{u \in U} \sum_{w \in W} e_{u,w},$$2 input power, with reported peak efficiencies of $$E_{\text{individual}, i}(\Delta t) = \sum_{u \in U} \sum_{s \in S} \Big(e_{u,s}\, t_{u,s}\Big) + \sum_{u \in U} \sum_{w \in W} e_{u,w},$$3 at $$E_{\text{individual}, i}(\Delta t) = \sum_{u \in U} \sum_{s \in S} \Big(e_{u,s}\, t_{u,s}\Big) + \sum_{u \in U} \sum_{w \in W} e_{u,w},$$4 and $$E_{\text{individual}, i}(\Delta t) = \sum_{u \in U} \sum_{s \in S} \Big(e_{u,s}\, t_{u,s}\Big) + \sum_{u \in U} \sum_{w \in W} e_{u,w},$$5 at $$E_{\text{individual}, i}(\Delta t) = \sum_{u \in U} \sum_{s \in S} \Big(e_{u,s}\, t_{u,s}\Big) + \sum_{u \in U} \sum_{w \in W} e_{u,w},$$6; the rectifier demonstrated operation from $$E_{\text{individual}, i}(\Delta t) = \sum_{u \in U} \sum_{s \in S} \Big(e_{u,s}\, t_{u,s}\Big) + \sum_{u \in U} \sum_{w \in W} e_{u,w},$$7 ($$E_{\text{individual}, i}(\Delta t) = \sum_{u \in U} \sum_{s \in S} \Big(e_{u,s}\, t_{u,s}\Big) + \sum_{u \in U} \sum_{w \in W} e_{u,w},$$8) to $$E_{\text{individual}, i}(\Delta t) = \sum_{u \in U} \sum_{s \in S} \Big(e_{u,s}\, t_{u,s}\Big) + \sum_{u \in U} \sum_{w \in W} e_{u,w},$$9 ($U = \{Pu, Su, Mu, TRu\}$0), with maximum rectifier PCE of about $U = \{Pu, Su, Mu, TRu\}$1 at $U = \{Pu, Su, Mu, TRu\}$2 (Martins et al., 2017). The RF-powered WPSN testbed instead prioritizes end-to-end system modeling: a Powercast P1110 rectenna feeds a $U = \{Pu, Su, Mu, TRu\}$3 supercapacitor with leakage resistance $U = \{Pu, Su, Mu, TRu\}$4, while the beacon adaptively sets amplifier duty cycle and output power to maintain a target stored energy of $U = \{Pu, Su, Mu, TRu\}$5 (Setiawan et al., 2017).

Storage selection is correspondingly diverse. Supercapacitors dominate batteryless and burst-driven designs because of high charge/discharge efficiency and fast transient response (Landivar et al., 2024, Setiawan et al., 2017). Hybrid battery–supercapacitor storage appears where communication bursts would otherwise stress the battery (Borza et al., 2019). In all cases, storage is not a passive reservoir but an active control variable: start thresholds, sleep thresholds, recharge windows, and capacitor voltage excursions determine when sensing and communication are admissible.

4. Sensing and communication co-design

Communication design is a primary determinant of whether a node remains autonomous. E-WAN addresses wide-area low-power energy-harvesting networks through three host-coordinated virtual sub-networks: a Multi-hop VSN using short-range synchronous FSK flooding, a Single-hop VSN using long-range LoRa TDMA rounds, and a Bootstrapping VSN using asynchronous long-range request/reply to deliver timing and scheduling information cheaply after depletion (Stricker et al., 30 Jun 2025). Nodes use simple local triggers—energy availability, receiving schedules, and missing schedules—to move among these VSNs. The architecture rejects the assumption that a single communication mode suffices: multi-hop is preferred when a path exists, but reliable single-hop is retained as a guaranteed fallback.

A different strategy is to suppress communication and sampling events altogether. The two-prong EAWSN built on VirtualSense combines always-on ultra-low-power wake-up receivers with Model-Based Sensing, in which a PIC16F1825 performs periodic sampling and derivative-based prediction without waking the main MCU (Raza et al., 2016). The radio wake-up receiver listens continuously at $U = \{Pu, Su, Mu, TRu\}$6 in the 868 MHz variant, while the MBS co-processor consumes $U = \{Pu, Su, Mu, TRu\}$7 active and $U = \{Pu, Su, Mu, TRu\}$8 in sleep (Raza et al., 2016). Because the node transmits only when the model tolerance is violated, traffic reduction reaches $U = \{Pu, Su, Mu, TRu\}$9 in the tunnel light dataset and $$S = \{\text{sleep}, \text{awake}, \text{active}, \text{idle}\}$$0, $$S = \{\text{sleep}, \text{awake}, \text{active}, \text{idle}\}$$1, and $$S = \{\text{sleep}, \text{awake}, \text{active}, \text{idle}\}$$2 for light, humidity, and temperature in the Intel dataset (Raza et al., 2016).

Protocol heterogeneity appears in batteryless indoor-light nodes as well. The BLE/LIoT design couples a BLE node based on the nRF52840 with a LIoT node using visible-light downlink and infrared uplink; both use bidirectional exchanges so that energy is not wasted on undelivered data (Landivar et al., 2024). The TDDC resistive-sensor node adopts another extreme: sensing itself is implemented as a timed capacitor discharge through the measurand,

$$S = \{\text{sleep}, \text{awake}, \text{active}, \text{idle}\}$$3

with the microcontroller counting timer pulses rather than sampling an analog voltage through a conventional ADC front-end (Costanza et al., 22 Jul 2025).

Delay analysis shows that sensing energy and transmission energy need not affect performance identically. In the harvest-then-use model with sensing cost $$S = \{\text{sleep}, \text{awake}, \text{active}, \text{idle}\}$$4, transmission cost $$S = \{\text{sleep}, \text{awake}, \text{active}, \text{idle}\}$$5, and retransmission window $$S = \{\text{sleep}, \text{awake}, \text{active}, \text{idle}\}$$6, the average update age does not depend on $$S = \{\text{sleep}, \text{awake}, \text{active}, \text{idle}\}$$7, whereas the average update cycle does (Liu et al., 2015). In the large-$$S = \{\text{sleep}, \text{awake}, \text{active}, \text{idle}\}$$8 limit,

$$S = \{\text{sleep}, \text{awake}, \text{active}, \text{idle}\}$$9

while

$$W = \{\text{is}, \text{sw}, \text{wa}, \text{ai}, \text{ia}\}$$0

(Liu et al., 2015). This formalizes a recurring EAWSN tension: more aggressive retransmission and update policies can improve update frequency while making delivered information older.

5. Representative empirical results

Reported EAWSN results are heterogeneous because different works optimize different subsystems: average node power, packets per joule, recharge time, delivered-packet periodicity, or control response. The following examples illustrate the operational range that has been demonstrated.

System	Representative result	Context
E-WAN (Stricker et al., 30 Jun 2025)	$$W = \{\text{is}, \text{sw}, \text{wa}, \text{ai}, \text{ia}\}$$1–$$W = \{\text{is}, \text{sw}, \text{wa}, \text{ai}, \text{ia}\}$$2 packets/J	15-node simulations; 20 B payload
WuR + DBP + MBS (Raza et al., 2016)	$$W = \{\text{is}, \text{sw}, \text{wa}, \text{ai}, \text{ia}\}$$3 and $$W = \{\text{is}, \text{sw}, \text{wa}, \text{ai}, \text{ia}\}$$4	Average node power in tunnel and Intel light cases
Batteryless BLE / LIoT (Landivar et al., 2024)	$$W = \{\text{is}, \text{sw}, \text{wa}, \text{ai}, \text{ia}\}$$5 and $$W = \{\text{is}, \text{sw}, \text{wa}, \text{ai}, \text{ia}\}$$6	Bidirectional exchange periods at $$W = \{\text{is}, \text{sw}, \text{wa}, \text{ai}, \text{ia}\}$$7 lx
Battery-free TDDC BLE sensor (Costanza et al., 22 Jul 2025)	$$W = \{\text{is}, \text{sw}, \text{wa}, \text{ai}, \text{ia}\}$$8	Resistance estimation using linear-fit mapping
RF harvester + UWB TX (Martins et al., 2017)	$$W = \{\text{is}, \text{sw}, \text{wa}, \text{ai}, \text{ia}\}$$9 at $$E_{\text{cap}}(t+1)=\max(\min(E_{\text{cap}}(t)+E_{\text{harv}}(t)-E_{\text{used}}(t),B),0),$$0; $$E_{\text{cap}}(t+1)=\max(\min(E_{\text{cap}}(t)+E_{\text{harv}}(t)-E_{\text{used}}(t),B),0),$$1 at $$E_{\text{cap}}(t+1)=\max(\min(E_{\text{cap}}(t)+E_{\text{harv}}(t)-E_{\text{used}}(t),B),0),$$2; TX $$E_{\text{cap}}(t+1)=\max(\min(E_{\text{cap}}(t)+E_{\text{harv}}(t)-E_{\text{used}}(t),B),0),$$3	Prototype power chain and sub-GHz UWB transmitter
Piezoelectric PID harvester (Kassan et al., 2019)	$$E_{\text{cap}}(t+1)=\max(\min(E_{\text{cap}}(t)+E_{\text{harv}}(t)-E_{\text{used}}(t),B),0),$$4 to $$E_{\text{cap}}(t+1)=\max(\min(E_{\text{cap}}(t)+E_{\text{harv}}(t)-E_{\text{used}}(t),B),0),$$5	Recharge level over the same harvest interval

At the network-protocol level, E-WAN roughly doubled efficiency relative to a single-hop-only baseline in the reported simulations, for example $$E_{\text{cap}}(t+1)=\max(\min(E_{\text{cap}}(t)+E_{\text{harv}}(t)-E_{\text{used}}(t),B),0),$$6 versus $$E_{\text{cap}}(t+1)=\max(\min(E_{\text{cap}}(t)+E_{\text{harv}}(t)-E_{\text{used}}(t),B),0),$$7 packets/J in the OB scenario with $$E_{\text{cap}}(t+1)=\max(\min(E_{\text{cap}}(t)+E_{\text{harv}}(t)-E_{\text{used}}(t),B),0),$$8, and maintained small downtime while providing a reliable single-hop fallback (Stricker et al., 30 Jun 2025). In the 12-hour indoor deployment, packets received were $$E_{\text{cap}}(t+1)=\max(\min(E_{\text{cap}}(t)+E_{\text{harv}}(t)-E_{\text{used}}(t),B),0),$$9 for node 1 and $B = 0.7\,\text{J}$0 for node 2, with simulated counterparts of $B = 0.7\,\text{J}$1 and $B = 0.7\,\text{J}$2 (Stricker et al., 30 Jun 2025). EDA’s own routing comparison similarly concluded that Selective routing yields longer lifetime than Random routing by shifting work from Global to Local and Individual constituents (Hoang et al., 2012).

At the node level, the wake-up-receiver/model-based-sensing platform reported more than three orders of magnitude reduction in average power relative to the baseline. In the tunnel case, the baseline was $B = 0.7\,\text{J}$3, the best wake-up-receiver-only configuration was $B = 0.7\,\text{J}$4, WuR plus DBP was $B = 0.7\,\text{J}$5, and WuR plus DBP plus MBS reached $B = 0.7\,\text{J}$6 (Raza et al., 2016). The BLE/LIoT batteryless nodes instead emphasize reliable end-to-end operation under indoor light: over 8-hour runs, BLE achieved packet delivery rates of $B = 0.7\,\text{J}$7 at $B = 0.7\,\text{J}$8 lx and $B = 0.7\,\text{J}$9 at $E_{\text{cap}}(t)=\frac{1}{2}CV_{\text{cap}}(t)^2$0 lx, while LIoT reported $E_{\text{cap}}(t)=\frac{1}{2}CV_{\text{cap}}(t)^2$1 at both light levels (Landivar et al., 2024).

Battery-free sensing results expose a different metric space. In the TDDC BLE node, a single measurement consumes

$E_{\text{cap}}(t)=\frac{1}{2}CV_{\text{cap}}(t)^2$2

while the send phase is sized for approximately $E_{\text{cap}}(t)=\frac{1}{2}CV_{\text{cap}}(t)^2$3; for a $E_{\text{cap}}(t)=\frac{1}{2}CV_{\text{cap}}(t)^2$4 test resistor, the measured discharge time was $E_{\text{cap}}(t)=\frac{1}{2}CV_{\text{cap}}(t)^2$5, corresponding to approximately $E_{\text{cap}}(t)=\frac{1}{2}CV_{\text{cap}}(t)^2$6 timer counts at $E_{\text{cap}}(t)=\frac{1}{2}CV_{\text{cap}}(t)^2$7 (Costanza et al., 22 Jul 2025). In the piezoelectric study, the principal empirical claim is not long-term uptime but energy-management benefit: with PID control, storage climbed to approximately $E_{\text{cap}}(t)=\frac{1}{2}CV_{\text{cap}}(t)^2$8 rather than approximately $E_{\text{cap}}(t)=\frac{1}{2}CV_{\text{cap}}(t)^2$9 over the same harvest interval (Kassan et al., 2019).

6. Trade-offs, misconceptions, and open problems

A recurring misconception is that EAWSN design can be solved by maximizing harvested energy alone. The literature instead treats autonomy as a cross-coupled problem involving constituent interactions, radio choices, storage dynamics, and workload shaping. EDA explicitly argues that reducing one energy constituent in isolation can increase others, and it identifies the absence of a single integrated total-energy formulation, explicit TX/RX/path-loss models, and a formal optimization problem as open issues (Hoang et al., 2012). E-WAN reaches a similar conclusion at the protocol level: local optimization toward multi-hop efficiency can shift relay burden onto other nodes, and globally energy-optimal scheduling remains difficult without accurate energy and link prediction (Stricker et al., 30 Jun 2025).

Another misconception is that battery-free operation eliminates the need for storage engineering. In practice, battery-free nodes are storage-centric. The indoor-light BLE/LIoT systems rely on a $C_{\text{stor}} = 440\,\mu\text{F}$00 supercapacitor and adapt sleep intervals to the harvested-power regime (Landivar et al., 2024). The TDDC sensor derives its entire measurement and send-phase budget from a $C_{\text{stor}} = 440\,\mu\text{F}$01 capacitor and is accuracy-limited by capacitor tolerance, GPIO output resistance, and threshold uncertainty; the paper reports model-based mapping with relative error within about $C_{\text{stor}} = 440\,\mu\text{F}$02, improved to within about $C_{\text{stor}} = 440\,\mu\text{F}$03 by a linear-fit calibration (Costanza et al., 22 Jul 2025). This suggests that autonomy and metrological accuracy may be constrained by different components.

Communication trade-offs remain particularly sharp. Multi-hop is often more energy-efficient than single-hop, but only when powered relays and a connected path exist; E-WAN therefore preserves long-range single-hop as a reliable fallback (Stricker et al., 30 Jun 2025). Wake-up receivers suppress idle listening, but deeper MCU sleep states introduce wake latency; the VirtualSense study reports $C_{\text{stor}} = 440\,\mu\text{F}$04 wake latency for standby and sleep, and $C_{\text{stor}} = 440\,\mu\text{F}$05 for hibernation (Raza et al., 2016). Optical communication can avoid RF congestion, yet the LIoT node requires much longer sleep intervals than the BLE node because active energy is $C_{\text{stor}} = 440\,\mu\text{F}$06 rather than $C_{\text{stor}} = 440\,\mu\text{F}$07 per cycle (Landivar et al., 2024). In the delay-analysis model, increasing retransmission window $C_{\text{stor}} = 440\,\mu\text{F}$08 improves update frequency but worsens delivered-update age (Liu et al., 2015).

Hardware-specific limitations are equally prominent. The piezoelectric PID study does not specify rectifier or converter implementations, storage chemistry, radio technology, or measured harvested power under real vibrations (Kassan et al., 2019). The RF-harvesting circuit work highlights a different boundary: higher $C_{\text{stor}} = 440\,\mu\text{F}$09 improves passive voltage boost but narrows bandwidth, and shared antennas complicate joint optimization of harvesting, reception, and transmission (Martins et al., 2017). The RF-powered WPSN requires a dedicated RF power beacon and operates under path-loss and EIRP constraints that sharply limit feasible range (Setiawan et al., 2017). EAWSN research therefore remains divided between architecture-level generality and platform-specific realizations.

Taken together, these works define the EAWSN not as a single canonical node but as a design regime in which harvested energy, storage state, sensing workload, and communication behavior are jointly managed so that operation can continue without battery replacement, and in some cases without a battery at all.