The smart watering brain (VRWE), in plain English
HomePrecisionThe smart watering brain (VRWE), in plain English
Precision · Smart watering

The smart watering brain (VRWE), in plain English

A grow room can water plants on its own by combining several sensor signals instead of trusting one moisture probe that might be lying.

Precision9 diagramsEvidence-linked · 5 sources~9 min read
In this guide
01 What this is, and the problem it solves02 Accuracy, self-review, and grain-of-salt notes03 Key terms, defined once04 Why fuse signals: the water bank account05 How trust and uncertainty work06 What it actually decides07 The shared-drain puzzle, and how it is untangled08 Pitfalls, and what fools a single sensor09 Realistic expectations
Start here

What this is, and the problem it solves

Grain of salt

Provisional: Multi-signal caution is sound engineering. Claims of never flooding or never starving depend on sensor health, calibration, and fail-safes — keep hard VWC floors and human override.

VRWE stands for Virtual Root-Zone Water Estimator: software that decides when and how much to water a plant by combining several signals instead of obeying one sensor. It is a ‘virtual’ estimator because it never reads the water directly. It works the amount out from several clues, the way you can tell a kettle is nearly empty from its weight and how long it has been boiling.

Each pot has only one moisture sensor, and that sensor feels only a tiny spot of soil, roughly the volume of a soda can[1]. If that spot happens to be dry, or if water sneaks past it, the sensor reports ‘I’m dry!’ and a dumb timer would believe it and drown the plant. VRWE treats the sensor as one opinion to double-check, not the boss.

How VRWE thinks: SEE, THINK, DO1SEEone sensor, atiny view thatcan lie→2THINKthe brain doeswater-balancemath→3DOwater a bit /wait / ask ahumanThe sensor is just the first step, not the decision.
Figure 1. VRWE sees one limited signal, thinks by checking it against other evidence, then acts conservatively. The whole paper is about the THINK box.
What kind of paper this is

This is operational, product-style guidance for how the system behaves, not a lab study. It pairs with the root-zone sensor paper (what a single probe actually measures) and the signal & noise paper (telling a real change from sensor jitter).

How sure is this?

Accuracy, self-review, and grain-of-salt notes

How sure is this paper?

We've gone to great lengths to keep these guides honest. One of the main ways we do that is self-review: we actively look for claims that are subjective, only lightly backed by literature, or based on grower practice rather than a controlled study — and we call those out instead of dressing them up as settled science.

Often there simply is no paper for the decision you're making. In those cases we're drawing on what other growers report and what has worked in our own rooms. That can still be useful — but it is not a lab proof. Do what works for your plants, your room, and your meters. If a table disagrees with your crop, believe the crop and log the difference.

Solid
Well supported by plant science, standards, or broad multi-source consensus
  • Single moisture probes can lie; multi-signal caution is sound engineering
Operational
What many growers and rooms actually run — start here, then tune
  • VRWE-style fusion as a safety architecture for automated irrigation
Grain of salt
Subjective, thin literature, single studies, or “this works for us” practice
  • Any claim the system 'never floods, never starves' in all failure modes
  • Transpiration proxies as precise water-need models without crop coefficients

See something glaringly wrong? Tell us and we'll fix it. Please open a GitHub issue with the paper name and what looks off (include a source if you have one): Report an accuracy issue. Local law, labels, and licences always override any recipe here. Inline notes labelled grain of salt flag the highest-risk over-trust points in the text.

Vocabulary

Key terms, defined once

Five words carry the whole idea, so we define them up front. Don’t memorise them. Each one comes back in context.

VWC (volumetric water content)How wet the soil is at the sensor’s spot, as a percentage. This is the raw ‘sensor reading’ VRWE double-checks.
Runoff / drainWater that leaves the bottom of the pot. Water that drains away never fed the plant.
Full pot (DUL, drained upper limit)The most water the pot can hold once it has finished dripping. Past this point, extra water just runs off.
ChannelingWater sneaking straight down one path and missing the roots. It goes in the top and out the bottom without doing any good.
ConfidenceThe brain’s self-rated trust in its own current guess. High confidence allows bolder action. Low confidence forces caution.
The five words that carry the whole idea.
TermPlain-English meaning
VWCHow wet the soil is at the sensor’s spot
Runoff / drainWater leaving the bottom of the pot
Full pot (DUL)The most water the pot holds once it stops dripping
ChannelingWater bypassing the roots straight to the drain
ConfidenceThe brain’s trust in its own estimate right now
A daily water-content cycleSaturate to field capacity, hold it, allow a controlled dryback, repeat. The size of the dryback is the steering lever.working bandtoo dry0255075100offP3 minonP1FCP2offwater content %
Diagram. The watering brain holds the daily water-content curve to setpoints instead of a fixed clock.
Saturation, field capacity and drybackthe same block at three points in a daySaturatedjust irrigated, almostno air→Field capacityfree water drained,daily peak→Dryback lowplant drank, air +oxygen in
Diagram. Full pot (drained upper limit) and the dryback low define the band it steers.
Why a dried-out block won’t just rewet from the dripperStill moist: rewets evenlywater spreads through the fibresToo dry: water channels pastdry coreruns down channels, core stays dryBelow ~30% WC the dry fibre matrix repels even spreading. Water finds preferential channels and exits as runoff, so the drip-rate that hydrated a wet block cannot re-saturate a dry one. You must hand-soak.
Diagram. Channeling: once a pot is too dry, water runs down preferential paths and the core stays dry.
Core idea 1

Why fuse signals: the water bank account

VRWE keeps a checkbook for water instead of believing one probe. Money IN is the water the drippers squirted, known precisely because drippers are calibrated, so you know exactly how much you put in. Money OUT is what the plant drank plus what drained away. The running balance is the water really in the pot.

The plant’s drinking, its transpiration, can be estimated from heat and light, because a plant pulls water faster when it is warmer and brighter[3]. So even without trusting the sensor, the brain has a good independent guess of OUT. The sensor becomes one statement to check against the balance, not the sole source of truth.

One watering cycle, as a bank balanceEach cycle: add what you poured in, subtract what the plant drank and what ran off.02142648555%Start75%+ IN (irrigation)66%- drank (uptake)60%- drained
Figure 2. The estimate is rebuilt every cycle like a running balance. The ending number is the brain’s best guess of real root-zone water, before it even looks at the sensor.
trusted
Inputs IN
Measured irrigation volume from calibrated drippers.
estimated
Inputs OUT
Plant uptake estimated from temperature and light, plus drainage.
cross-check
One vote
The raw sensor VWC reading, checked, not obeyed.
output
Estimate + confidence
All signals combined into one number, with a trust score attached.
Figure 3. Several signals feed one combined estimate. Because the brain has multiple independent clues, a single wrong signal can be outvoted instead of obeyed.
The point of fusing signals

Listen to only one sensor and a single bad reading becomes a bad decision. When several independent signals all feed the estimate, one liar gets outvoted. The system stays right even when one input is wrong.

Core idea 2

How trust and uncertainty work

The brain carries a confidence meter alongside every estimate, answering ‘how sure am I?’. A number on its own (‘58% wet’) tells you nothing about whether to bet on it.

Confidence is high when the independent signals agree, when the sensor, the bank balance and the uptake estimate all point the same way. Confidence drops when they disagree, when the sensor says dry but the bank balance says the pot is full. A lying sensor can only make the system more cautious. It can never trick the brain into believing there is more room to add water than the balance allows, and that is what protects the plant.

The confidence meter and what each zone unlocksHigher confidence unlocks bolder action; low confidence forces caution.Low: wait / askMedium: small safe sipHigh: water normally0%50%100%
Figure 4. Confidence is a dial, not a yes/no. High confidence lets the brain water a full measured amount. Medium allows only a small safe sip. Low means wait or ask a human.
A faulty sensor is safe
  • Every estimate ships with a confidence level, not just a number.
  • Agreement between independent signals raises confidence; disagreement lowers it.
  • Low confidence triggers caution, never bold action.
  • A faulty sensor biases toward ‘wait’, never toward flooding or starving.
Core idea 3

What it actually decides

The brain only ever picks one of three outcomes, driven by confidence and headroom (how much room is left before the pot is full).

The three possible decisions1Confident + room to fillwater a measuredamount→2Unsurewait, or delivera small safe sip→3Stuckescalate to ahuman instead ofguessingEvery path obeys one bound: prefer temporary mild deficit over flooding when uncertain; hard-floor emergency VWC still required.
Figure 5. The whole decision logic collapses to three branches, and all three are bounded by the same promise.

It waters a bit when the brain is confident and there is room to fill. It waits or gives a small safe sip when it is not sure, rather than committing to a full shot. It asks a human when it is genuinely stuck, when the signals contradict each other and it cannot resolve them. The whole logic is fenced in by one rule: prefer temporary mild deficit over flooding when uncertain; hard-floor emergency VWC still required.

The golden rule

Water more only when confident. When in doubt, do the safe thing. That single bound is what turns ‘automatic watering’ from a scary idea into a safe one.

In practice

The shared-drain puzzle, and how it is untangled

A real install is messier than one pot. Often three grow rooms drain into one shared sump (a collection bucket with a pump), and the air conditioner and dehumidifier drip into that same bucket. When the pump flushes, who caused it is unclear: a plant overflowing, or just the AC condensate. VRWE untangles this in three steps.

  1. 1
    Learn the background drip
    At night, when irrigation is off, the only water reaching the sump is the AC and dehumidifier condensate. The brain learns that steady background drip and subtracts it from every later reading.
  2. 2
    Stagger the watering times
    Each room waters at a different time. Because the flush timing now lines up with one room’s watering, the brain can tell which room caused each flush.
  3. 3
    If still ambiguous, say so
    If two events overlap and attribution is genuinely unclear, the reading is marked ‘not sure’ rather than guessed. The same fail-safe instinct applies as everywhere else.
source
Room A
Drains into the shared sump.
source
Room B
Drains into the shared sump.
source
Room C
Drains into the shared sump.
noise
AC + dehumidifier
Constant condensate drip, learned and subtracted at night.
puzzle
Shared sump + pump
One bucket, one flush. Who caused it?
fix
Untangle
Subtract background, stagger timing, flag the unclear ones.
Figure 6. Three rooms plus the AC and dehumidifier all empty into one sump, so a single flush is ambiguous. The three untangling steps recover which room caused which flush.
Staggered watering makes each flush traceableThe low flat baseline is the AC drip; each spike sits under the room that just watered.01225385000:00A waters+30mB startB waters+30mC startC waters+30mdrain rate
Figure 7. Because the rooms water at different times, each drain spike falls directly under the room that caused it. A flush right after a room waters means that room is full or overflowing.
What a post-watering flush tells you

A flush right after a given room waters is a clear read-back: that room reached ‘full pot’ and the extra ran off. That is useful information, not a fault.

Watch out

Pitfalls, and what fools a single sensor

The failure modes below are the situations VRWE is built to survive. They are the reason it exists. The defence is the same in every case: cross-check against the water balance and drop confidence, rather than acting on a lone suspicious reading.

fooled
Dry-pocket sensor
The probe sits in a local dry spot and reads dry while the rest of the pot is fine.
fooled
Channeling
Water races down one side, past both sensor and roots, straight to the drain.
fooled
Shared-drain ambiguity
One bucket hides which room actually overflowed.
Figure 8. Three classic ways a single signal lies. None of them fools VRWE, because the balance and the confidence meter catch the contradiction.

Channeling is worth a closer look, because it is sneaky. In container substrate, water can follow a preferential flow path, a fast channel that routes irrigation past the root zone entirely[4]. The pour-in volume looks healthy, but the water never reaches the roots. It just shows up as drain. How quickly water moves through and out of a soilless mix depends on the substrate’s own physics[5], which is why the brain watches drain timing, not just drain volume.

in
Top of pot
Irrigation enters here.
bypass
Channel path
Water shoots down one side, past the sensor and roots.
out
Drain
Comes out the bottom almost immediately. Roots stayed dry.
Figure 9. A cross-section of channeling: water in the top, out the bottom, roots untouched. To the balance this looks like ‘lots in, lots straight out’, a tell-tale the brain uses to lower confidence.
The defence is always the same

VRWE does not obey a reading that looks suspicious. It reconciles against the bank balance, and if they disagree it lowers confidence and acts cautiously. A single fooled sensor never becomes a flooded or starved plant.

Reality check

Realistic expectations

What VRWE promises, and what it does not
  1. The golden rule. It waters more only when confident; otherwise it does the safe thing. It is a safety-first estimator, not a mind reader.
  2. Worst case is over-caution. A bad sensor makes it cautious, not catastrophic. It will pause or ask before it ever floods or starves.
  3. Expect ‘wait’ and ‘ask a human’ by design. Those are the system working, not failing.
  4. Garbage in, garbage out. The estimate is only as good as its inputs. Accurate dripper volumes and a learned drain baseline matter. Sensor calibration drift quietly erodes every estimate that depends on it.[2]

VRWE trades a little speed for a lot of safety. It will occasionally hold back when a dumb timer would have charged ahead, and that is exactly the point. To go deeper on what a single probe really measures, read the root-zone sensor paper. To understand how the brain tells a real change from sensor noise before it ever acts, read signal & noise.

Related papers
Root-zone state · TEROS-12
What the sensor really sees
Signal & noise
Real signal vs sensor noise
The closed loop
Levers, signal & plant state

References

  1. Szerement, J., Woszczyk, A., Szypłowska, A., Kafarski, M., Lewandowski, A., Wilczek, A., & Skierucha, W. (2019). A Seven-Rod Dielectric Sensor for Determination of Soil Moisture in Well-Defined Sample Volumes. Sensors, 19(7), 1646. https://doi.org/10.3390/s19071646
  2. Mane, S., Das, N., Singh, G., Cosh, M., & Dong, Y. (2024). Advancements in dielectric soil moisture sensor calibration: A comprehensive review of methods and techniques. Computers and Electronics in Agriculture, 218, 108686. https://doi.org/10.1016/j.compag.2024.108686
  3. Koehler, T., Wankmüller, F. J. P., Sadok, W., & Carminati, A. (2023). Transpiration response to soil drying versus increasing vapor pressure deficit in crops: physical and physiological mechanisms and key plant traits. Journal of Experimental Botany, 74(16), 4789-4807. https://doi.org/10.1093/jxb/erad221
  4. Owen, J., & Norden, D. (Profile Products). Understanding drainage in horticultural growing media. Greenhouse Management. (industry/manufacturer or non-journal source) https://www.greenhousemag.com/article/growing-media-defining-drainage-improve-substrate/
  5. International Society for Horticultural Science (ISHS). Utilizing the HYDRUS model as a tool for understanding soilless substrate water dynamics. Acta Horticulturae 1168. https://www.ishs.org/ishs-article/1168_41

Citations marked in-text as [n] map to this list. Primary literature and official guidance except where noted. Cannabis tissue culture is strongly genotype-dependent, verify dilutions, hormone doses and local regulations against the primary sources before relying on them.