---
title: 'Internet of Plants: Biohybrid Networks'
url: https://www.emergentmind.com/topics/internet-of-plants-iop
type: topic
---

# Internet of Plants: Biohybrid Networks

Internet of Plants (IoP) denotes an emerging paradigm in which plants are treated as interconnected nodes within ecological and technological networks, with plant signaling analyzed using information and communication technology frameworks and plant-centered sensing, connectivity, analytics, and actuation organized around plant state rather than generic device telemetry [2509.08434]. In current research, IoP encompasses communication-theoretic models of native plant signaling, biohybrid architectures that embed engineered “agent plants” within plant communities, and deployed systems for monitoring, irrigation, greenhouse control, disease detection, and human-plant interaction [2406.11174].

## 1. Scope, definition, and acronym ambiguity

Within the plant literature, IoP is not merely a synonym for agricultural IoT. The stronger formulation treats plants as communicative entities whose exchanges can be described through transmitters, channels, receivers, noise, interference, and network topology, while also allowing those biological networks to interface with cloud services, mobile applications, AI models, and edge infrastructure [2509.08434]. In this sense, IoP spans both naturally occurring signaling networks and cyber-physical systems that monitor, interpret, or augment those networks.

The acronym is, however, ambiguous across fields. One paper uses IoP to mean “Internet of Production,” specifically an electric power SCADA environment within industrial cyber-physical systems rather than anything botanical [2312.15319]. Another uses IoP to mean “Internet of People,” in an edge-computing architecture for ambience intelligence rather than plant systems [2011.00236]. That ambiguity matters because some architectural or security ideas from those literatures may be transferable, but they do not define the botanical IoP domain itself. In the plant-centered sense, IoP refers to living plant networks and to digital infrastructures built around plant communication, plant monitoring, or plant-specific decision support.

A second distinction concerns what counts as “plant-centered.” Some systems monitor environmental proxies near a plant, such as soil moisture, humidity, light, or water level, and thereby construct an indirect but actionable representation of plant state. Others attempt to work closer to native plant signaling, using volatile organic compounds (VOCs), electrical activity, mycorrhizal connectivity, or stress acoustics as the communication substrate itself [2509.08434]. This suggests that IoP is best understood as a spectrum: at one end are plant-aware IoT systems, and at the other are biohybrid communication networks in which the plant network is itself part of the sensing and signaling medium.

## 2. Native signaling modalities and ICT abstractions

The communication-theoretic foundation of IoP is multimodal. The literature consistently identifies four major signaling classes: chemical signaling aboveground and belowground, mycorrhizal-network-mediated signaling, electrical signaling, and acoustic signaling [2509.08434]. Aboveground chemical communication is typically modeled through VOC emission, advection, diffusion, and uptake; belowground chemical communication adds dual-phase soil transport, sorption, microbial degradation, and long memory. Mycorrhizal communication is treated as signaling through common mycorrhizal networks (CMNs), while electrical signaling is framed around action potentials, variation potentials, and system potentials propagating through excitable tissues or conductive couplings. Acoustic signaling is the least mature channel, but plant stress emissions and plant responses to sound are now sufficiently documented to be included in IoP formulations.

In ICT terms, these modalities map cleanly onto different channel classes. VOC signaling resembles molecular communication, with concentration or blend ratios serving as candidate encodings, substantial intersymbol interference, and strong environmental dependence. The end-to-end odor communication model based on Green Leaf Volatiles (GLVs) makes this explicit by modeling airborne transport through a time-varying diffusion-advection channel,
$$
\frac{\partial c}{\partial t}(\mathbf{x}, t) = D \nabla^{2} c(\mathbf{x}, t) - \mathbf{v}(t) \cdot \nabla c(\mathbf{x}, t) + S(\mathbf{x}, t),
$$
with a damaged transmitter plant emitting \((Z)\)-3-hexenal, \((Z)\)-3-hexenol, and \((Z)\)-3-hexenyl acetate, and a sink plant decoding them biochemically through the defensive metabolite \((Z)\)-3-hexenyl \(\beta\)-vicianoside [2602.16850]. A key result of that model is that \((Z)\)-3-hexenol is the primary driver of the system and that plant perception generally operates in a non-linear region [2602.16850].

Mycorrhizal signaling is often reformulated as a graph-structured shared backbone rather than a point-to-point link. A representative network model is
$$
\dot{\mathbf{C}}(t) = -K\mathbf{L}\,\mathbf{C}(t),
$$
where \(\mathbf{L}\) is the graph Laplacian of the fungal network and \(K\) is an effective hyphal conductance [2509.08434]. This abstraction is important because it moves IoP beyond isolated plant pairs and toward network effects such as nutrient redistribution, warning propagation, and topology-dependent delay. Electrical signaling, by contrast, is the fastest established plant channel and is commonly represented by passive cable or Hodgkin–Huxley-type dynamics; the passive form
$$
\frac{1}{r_i}\,\frac{\partial^2 V}{\partial x^2} - \frac{V}{r_m} - C_m\,\frac{\partial V}{\partial t} = 0
$$
captures membrane-voltage propagation with explicit axial resistance, membrane resistance, and capacitance [2509.08434].

Acoustic signaling occupies a more tentative position. The survey literature treats stress acoustics as event-driven impulsive emissions, often linked to cavitation, with propagation through air, tissue, or soil and potential reception through mechanosensitive channels or organ-level mechanical sensing [2509.08434]. However, it also states clearly that definitive evidence of plant-to-plant acoustic communication remains lacking [2509.08434]. This distinction is important: acoustic emissions are empirically useful as stress indicators, but the communicative interpretation is still under consolidation.

## 3. Architectural patterns and system design

Implemented IoP systems show a recurrent layered structure even when their technologies differ. Precision-agriculture systems have been organized into field, fog-computing, cloud-computing, and data-presentation layers, with local summarization and event generation used to reduce communication and energy cost [1906.02678]. Mobile/cloud plant-care systems have been organized into front end, back-end storage and data retrieval, sensors, and chatbot, with ThingSpeak for telemetry ingestion, Firebase-style real-time and NoSQL storage, Flutter for the client, and Gemini for natural-language interpretation of plant telemetry [2409.15910]. Low-cost irrigation platforms adopt a three-tier sensing layer, processing layer, and application layer, typically centered on an ESP32 or similar microcontroller, local displays and alerts, and a cloud platform such as ThingSpeak for remote visibility and historical analysis [2601.15830].

These architectures encode different answers to the same design problem: how to convert plant-relevant signals into persistent, queryable, and actionable state. In field deployments, the emphasis is often on data reduction and robust transport. In domestic or greenhouse systems, the emphasis shifts toward immediate actuation, local human interaction, and cloud dashboards. In conversational systems, the plant’s sensed state becomes a semantic object. The LLM-based plant communication prototype is especially explicit on this point: telemetry is transformed into the plant’s “health,” “mood,” and care recommendations, with species-specific prompting such as “Imagine you are a cactus….” rather than model fine-tuning [2409.15910]. The result is an IoP architecture in which the plant is represented as an interactive digital agent grounded in sensor-derived context.

At the speculative end of the field, the “agent plant” architecture proposes a more radical biohybrid design. An agent plant is defined as an artificially created, plant-like structure embedded within plant fields whose tasks include communicating with surrounding plants, reporting to a central agent plant, constructing plant–agent “highways” for chemical and electrical signals, storing nutrients, drugs, and infochemicals, and imitating mycorrhizal fungi–plant symbiosis for energy harvesting [2406.11174]. Its units are a processor, communication unit, storage unit, and power unit. No implementation-level signaling protocol is given, but the architecture is notable because it treats communication, intervention, and powering as one integrated IoP problem rather than separate subsystems.

## 4. Implemented IoP systems

The most mature applied strand of IoP is plant-environment monitoring with local or cloud-mediated interpretation. In distributed precision agriculture, field nodes measuring air temperature, air humidity, solar radiation, soil temperature at various depths, wind speed, and rainfall have been coupled to fog nodes that perform 30-minute batching, min/max/mean extraction, outlier/event checks, trend-change detection, and symbolic aggregation using PAA and SAX before sending compact representations upward [1906.02678]. Cloud-side reconstruction then uses interpolation, with linear interpolation yielding RMSE \(= 0.0795\) for soil temperature and \(0.2551\) for solar radiation, while pchip is argued to be more physically plausible because it preserves monotonicity and avoids overshoot [1906.02678]. The contribution is architectural rather than agronomic: local intelligence close to the field reduces communication load while retaining plant-relevant temporal structure.

Domestic IoP systems emphasize plant-vicinity sensing and human-in-the-loop care. A winter-survival prototype for indoor plants used a Ruuvitag for temperature and relative humidity, a smartphone light sensor for illuminance, Bluetooth for local communication, SQLite and CSV logging, and species-specific thresholds stored for each plant profile [2106.05130]. In a 14-day Peace Lily experiment, average conditions differed sharply by micro-location: 19.48 °C, 34.24 humidity, and 10.36 lux in a corner versus 17.59 °C, 35.86 humidity, and 75.55 lux near a window, demonstrating that room-level averages can obscure the plant’s actual microclimate [2106.05130]. Another home-garden prototype combined soil moisture, DHT11 temperature/humidity sensing, RGB leaf-color sensing, and PIR-based security into a Blynk-connected system with automatic, scheduled, and manual watering and a rule-based health score of 30%, 60%, or 90% depending on whether one, two, or three factors pass the thresholds [2202.08153].

Controlled-environment agriculture extends the same logic into hydroponic and aeroponic settings. A hydroponic greenhouse prototype built around Arduino Mega, ESP8266, ThingSpeak, DHT-22, DS18B20, PH4502C, LDR, relays, and submersible pumps monitored water pH, light, water temperature, greenhouse temperature, and humidity and partially automated pump activation when pH or temperature reached certain levels [2305.01189]. The reported adequate-growth ranges were pH \(6.5\) to \(8\), greenhouse temperature \(26^\circ\mathrm{C}\) to \(29^\circ\mathrm{C}\), water temperature \(28^\circ\mathrm{C}\) to \(31^\circ\mathrm{C}\), and humidity \(70\%\), but the prototype also exposed practical calibration issues, including negative pH values caused by improper calibration [2305.01189]. A later aeroponic greenhouse for Geranium integrated SHT75, BH1750, TCS3200, YF-S201, and SRF05 sensing with Ubidots-based remote monitoring and CNN-based classification of healthy, drought-stressed, and rust-infected leaves [2509.12274]. On industrial-greenhouse training data, VGG19 reached \(0.9294\) accuracy, outperforming InceptionResNetV2 and InceptionV3, and on a 798-image experimental-greenhouse dataset VGG19 achieved \(86.34\%\) overall accuracy, with healthy leaves easiest and drought stress hardest to classify [2509.12274].

Low-cost irrigation platforms provide a more classical IoT formulation of IoP. An ESP32-based system with DHT22, capacitive soil moisture sensing, HC-SR04 water-level measurement, optional DS18B20 nutrient-solution sensing, OLED feedback, buzzer, RGB LED, ThingSpeak upload, and pump actuation maintained soil moisture within the optimal range of 60–80% field capacity 92% of the time, reduced water consumption by approximately 40% compared to manual irrigation, achieved \(99.7\%\) data-upload success, and delivered 156 of 156 alerts successfully [2601.15830]. The architecture remains threshold-based, but it demonstrates that even relatively simple plant-centered telemetry, actuation, and cloud analytics can materially improve water-use efficiency.

The most distinctive recent application is advanced human-plant interaction. In the LLM-mediated prototype, the actual testbed used a soil-moisture sensor and a DHT11 sensor, uploaded telemetry through ThingSpeak and a real-time database, and rendered the plant’s condition in first-person language via Gemini, with examples such as “I’m feeling a bit thirsty today, could you water me?” or “Greetings I am thriving in this cozy environment” [2409.15910]. Its novelty lies less in new sensing hardware than in closing the loop between telemetry, cloud-connected storage, species-aware prompting, and conversational UI. That system makes explicit a broader tendency within IoP: the shift from raw environmental dashboards toward plant-centered semantic interfaces.

## 5. Infrastructure, reliability, and security constraints

Plant-centered intelligence depends on data plumbing that is often more difficult than sensing itself. The ORPHEUS living lab addresses this layer directly through a three-layer sensor-node, gateway, and backend architecture using LoRa or BLE at the edge, gateway-side local buffering, secure TCP/IP upload with TLS/SSL when internet is available, ChirpStack and InfluxDB for LoRaWAN deployments, MongoDB for LoRa data, and Grafana and Django for visualization and querying [2111.09422]. The custom sensor nodes are designed to be rugged and energy-saving, with IP67-targeted enclosures, HDC2010 temperature/humidity sensing, support for TEROS 12 soil probes and in-house nitrate sensors, and a reported lifetime of 3–5 months on 4 AAA batteries [2111.09422]. This infrastructure work is significant for IoP because it shows that plant-aware systems require robust transport, persistence, metadata, and monitoring services before higher-level inference becomes operationally useful.

The applied IoP papers also reveal recurrent reliability limitations. The indoor winter-survival system required the Android smartphone to remain powered, nearby, and actively running, and the application crashed multiple times during the 14-day experiment, causing significant gaps in the dataset [2106.05130]. The LLM-based plant communication prototype does not specify lower-level protocols, explicit latency measurements, fallback behavior, or LLM safeguards, and it minimally addresses encryption in transit, database access policies, sensor spoofing, API-key security, prompt injection, or cloud-stored data governance [2409.15910]. The hydroponic greenhouse system exposed improper pH calibration and described the humidity sensor as not reliable compared to a commercialized device [2305.01189]. The aeroponic greenhouse, while stronger on plant-state inference, leaves exact environmental setpoints, control hysteresis, pH/EC management, CO\(_2\) monitoring, and fault-tolerance details unspecified [2509.12274]. These are not incidental issues; they indicate that much of IoP remains at prototype or pilot scale rather than at hardened infrastructure scale.

Cybersecurity is a particularly underdeveloped dimension in the botanical IoP literature. Threat modeling work that explicitly uses the acronym IoP is often about the “Internet of Production,” not the Internet of Plants, and therefore analyzes SCADA, RTUs, PLCs, and cloud-connected industrial control assets rather than botanical systems [2312.15319]. A plausible implication is that plant-specific IoP security has not yet received the same formal treatment as other cyber-physical domains. Existing plant papers discuss authentication, cloud services, or transport security only unevenly, and rarely model adversarial manipulation of sensors, models, or actuation.

## 6. Research frontier and unresolved questions

The research frontier is increasingly defined by end-to-end multimodal modeling. The tutorial literature argues that IoP needs integrated transmitter–channel–receiver models for chemical, electrical, mycorrhizal, and acoustic signaling, together with better non-invasive sensing, benchmark datasets, and AI methods that can interpret noisy multimodal plant data under realistic environmental variability [2509.08434]. The GLV communication model already points in that direction by treating plant-to-plant odor signaling as a complete communication system with a transmitter, time-varying physical channel, biochemical receiver, and biologically meaningful alarm variable. Its conclusion that \((Z)\)-3-hexenol dominates the response and that the receiver typically operates in a non-linear regime implies that IoP cannot safely assume linear, stationary plant channels [2602.16850].

A second frontier is spatial inference. The multimodal stress-source localization study treats a canopy as a bio-hybrid sensing network of “agent plants” instrumented for acoustic and VOC sensing and proposes a two-stage pipeline: coarse localization by acoustic time-difference-of-arrival multilateration, followed by VOC-based local refinement [2606.16618]. On a 52-scenario open-source dataset, TDOA multilateration achieved sub-meter mean absolute error once three or more agent plants were within acoustic range, whereas VOC-only localization remained above 3 m MAE at all tested densities [2606.16618]. Fusion usually stayed very close to TDOA-only, which led the authors to characterize acoustic localization as deployable with current hardware and VOC localization as a forward-looking capability pending compact biochemical sensors [2606.16618]. This is a substantial shift in IoP emphasis: from merely detecting stress to locating its source within a plant network.

Energy autonomy remains more speculative. The agent-plant proposal suggests that an engineered node could imitate mycorrhizal fungi–plant symbiosis, extract glucose through underground exchange, and feed it to a respiration-based biocell with anode, cathode, and nanochannels [2406.11174]. Its analytical core models expected extracted glucose as
$$
E[G_i] = n \cdot g_i \cdot p,
$$
with \(n\) connected plants, \(g_i\) glucose extracted per event, and \(p\) extraction probability, and relates expected power density to a Michaelis–Menten-like saturation in anode glucose concentration [2406.11174]. The cited respiration-based biocell studies report current densities up to \(6.42\,\mathrm{mA\,cm^{-2}}\) and maximum power density up to \(1.21\,\mathrm{mW\,cm^{-2}}\), but the plant-side extraction mechanism itself remains conceptual, and the full energy budget of an operational agent plant is not yet demonstrated [2406.11174].

The applied systems literature also points toward richer IoP stacks. Several papers explicitly or implicitly motivate multimodal sensing beyond moisture and climate, including pH, nutrients, light intensity, disease imaging, and possibly direct plant phenotyping; tighter integration of cloud dashboards with local actuation; and edge/cloud collaboration for lower latency and better privacy [2409.15910]. A plausible implication is that future IoP systems will increasingly resemble digital plant infrastructures rather than isolated monitoring gadgets: each plant or cultivation zone would combine live telemetry, species-specific requirements, historical trends, plant-state inference, and intervention logic within one persistent computational representation. The main unresolved question is whether those future systems will remain plant-aware IoT platforms, or whether they will mature into true biohybrid communication networks that operate through the plants’ own signaling modalities as first-class channels.

Source: https://www.emergentmind.com/topics/internet-of-plants-iop