Connected Stable Monitoring Systems

What connected stable monitoring means

A connected stable monitoring system is a combination of sensors, communications equipment and software used to collect, transmit, store and present information about horses or their environment. The system may be as simple as a network camera with a mobile app, or as complex as a multi-stable installation combining video, temperature and humidity sensors, water-flow meters, door contacts, microphones, wearable accelerometers, gateway hardware and cloud analytics.

The important distinction is between monitoring and control. Monitoring systems measure or record conditions and activity. Control systems can change conditions, for example by operating ventilation, lighting, doors or feeding equipment. A monitoring system may trigger an alert, but the alert is not itself an assessment of welfare or a diagnosis.

For a horse owner, the practical purpose is usually one or more of the following:

• extending observation beyond scheduled yard visits;
• checking that routine activities such as drinking, eating, resting or movement appear normal for an individual horse;
• identifying changes in stable conditions, including excessive heat, cold, humidity, water interruption or power failure;
• supporting staff coordination and handover on a commercial yard;
• creating a time-stamped record of an event or management change; and
• building an individual baseline against which unusual changes can be recognised.

These objectives should be defined before equipment is selected. A camera designed to show whether a foal is standing or lying is not automatically suitable for measuring respiratory rate. A wearable designed to estimate activity in a field is not necessarily appropriate for a horse wearing rugs, tack or stable equipment. Product descriptions often use broad terms such as “health monitoring”, while the underlying measurement may be no more than movement, temperature or location.

In Great Britain, technology must be understood as an aid to the responsibilities already placed on the person responsible for the horse. The Defra code of practice describes duties including providing a suitable environment, suitable diet, appropriate company, the ability to behave normally and protection from pain, suffering, injury and disease. The code is guidance rather than a technical specification for sensors, and it does not turn remote monitoring into a substitute for direct care. Defra, Code of practice for the welfare of horses, ponies, donkeys and their hybrids.

System architecture

Most connected systems contain five functional layers.

1. Measurement layer: cameras, microphones, thermometers, hygrometers, accelerometers, GPS units, door contacts, light sensors, pressure sensors or water meters.
2. Edge or gateway layer: a local device that receives data from sensors, performs initial processing and forwards information through Wi-Fi, Ethernet, Bluetooth, cellular or another radio protocol.
3. Connectivity layer: the yard’s broadband, local network or mobile network. Rural connectivity is often the limiting factor rather than the sensor itself.
4. Application layer: a mobile app, web dashboard, local display or alerting service.
5. Data and decision layer: historical records, trend analysis, algorithms, user-defined thresholds and notifications.

Some systems are cloud-dependent: the camera or sensor sends information to a remote service, and the user accesses it through the internet. Others process more data locally and continue to record during an internet outage. This distinction is central to resilience. A system that cannot retain footage or generate a local alert when broadband fails may provide less protection than its marketing suggests.

Protocols and communications

Wi-Fi is convenient and supports high-bandwidth video, but coverage can be unreliable through dense masonry, metal cladding and long stable blocks. Bluetooth Low Energy is efficient for nearby wearables and battery-powered sensors but normally requires a gateway. Cellular devices avoid dependence on the yard’s broadband but introduce signal, subscription and power-consumption constraints. LoRaWAN and similar low-power wide-area technologies can carry small amounts of sensor data over long distances, making them useful for environmental or gate sensors, but they are not a general replacement for video.

Manufacturers should state whether the device records during a connection failure, how data are buffered, what happens when a gateway loses power and how alerts are prioritised after reconnection. A “connected” product with no documented offline behaviour is difficult to risk-assess.

Types of monitoring

Video and still-image monitoring

Video is the most intuitive form of remote observation. It can help a responsible person see posture, gross movement, interaction with the stable environment, whether a horse is trapped or whether an obvious event has occurred. Night vision, infrared illumination and low-light sensitivity are relevant, but infrared images do not provide a reliable clinical temperature measurement. A thermal camera may show surface temperature patterns, yet interpretation is affected by coat, rugs, wetness, distance, emissivity and ambient conditions.

Camera placement is a construction and safety issue as well as an imaging issue. Equipment should be securely mounted, protected from chewing and impact, and positioned so that cables cannot become accessible to the horse. The UK government’s advisory standards for animals used in science specifically emphasise that cabling should be secured away from the equine and should not create a health risk. The same engineering principle is sensible in ordinary yards. UK government advisory standards for the care and accommodation of animals in science.

Video analytics may classify behaviours such as standing, lying, walking, rolling, eating or prolonged inactivity. These classifications are model outputs, not direct observations of welfare. A horse lying down can be normal; prolonged recumbency, repeated attempts to rise, or absence of expected behaviour may warrant human attention, but the meaning depends on the individual, time of day, housing, weather, management and other evidence.

Environmental monitoring

Environmental sensors commonly measure air temperature, relative humidity, light, noise, carbon dioxide or particulate matter. In some yards, water-flow sensors, tank-level sensors, smoke detectors, power monitors and door contacts are more operationally valuable than horse-worn devices.

Temperature and humidity sensors should be specified by measurement range, stated accuracy, resolution, response time, calibration method and installation position. A sensor placed above a horse’s back, next to a roof sheet or directly in a draught may not represent the air actually experienced by the horse. Relative humidity is temperature-dependent, and a single humidity value should not be treated as a complete description of stable air quality.

Carbon dioxide can be useful as an indicator of ventilation performance and occupancy-related accumulation, but it is not a complete measurement of air quality. Dust, ammonia, mould spores and other contaminants require different methods. Equally, a satisfactory sensor reading at one point in a stable does not prove that every box, corner or low-level air space has equivalent conditions.

Wearable activity and location sensors

Wearables typically use accelerometers, gyroscopes, magnetometers, GPS or combinations of these. An accelerometer measures changes in motion along one or more axes. Algorithms can convert these signals into estimates such as activity counts, lying bouts, movement periods or broad behavioural categories. GPS provides position and sometimes speed or distance, but its accuracy is affected by satellite visibility, antenna position, buildings and update interval.

Equine research supports the potential of accelerometers, but also illustrates why validation matters. A peer-reviewed study on adult domestic horses compared triaxial accelerometer data with live observation and video for lying behaviour. It found that lying events occurred overnight in the monitored horses, while also demonstrating the need to validate the device and attachment method for equines rather than simply assuming that livestock technology transfers directly. Validation of triaxial accelerometers to measure the lying behaviour of adult domestic horses.

More recent work has examined collars, halters and solar-powered GPS and acceleration devices for group-housed horses. Such studies are promising, but research prototypes, small study populations and limited monitoring periods should not be confused with universal clinical or welfare validation. A 2026 case study described different device designs, including a research device capable of collecting higher-frequency raw data and a lower-maintenance tag-like device transmitting limited daily data. The difference illustrates a general trade-off: richer data require more power, storage, bandwidth and analytical support. Suitability of solar-powered acceleration and global positioning devices for remote monitoring of equine welfare.

Water, feed and access monitoring

Water meters, tank-level sensors and automatic drinker monitors can identify interruption or unusual use. They measure access or flow, not necessarily ingestion. A horse may activate a drinker without consuming an adequate quantity, or may obtain water elsewhere. Similarly, a feed hopper may record a dispensing event without proving that the horse ate the ration.

Access monitoring is particularly relevant on group yards. The system needs to distinguish between the presence of an animal and the individual’s actual access to feed or water. Radio-frequency identification, visual recognition or location tags may assist, but all have failure modes. Bullying, blocked access, sensor occlusion and shared resources can produce misleading records.

What data can and cannot tell you

The strongest use of monitoring data is often comparative rather than absolute. An individual baseline may show that a horse normally rests during a particular period, moves frequently overnight or drinks at fairly consistent intervals. A sustained departure from that pattern can justify a prompt physical check.

Baseline comparison is more defensible than relying on a universal threshold because horses differ by age, temperament, workload, turnout regime, stable design, season and social circumstances. The baseline must also be refreshed when management changes. A horse moved from individual stabling to group turnout may show a large activity change that is expected rather than alarming.

False positives are inevitable. A camera may mistake bedding movement for a horse; an accelerometer may interpret a rug adjustment as activity; a GPS unit may show an implausible jump in position; an environmental sensor may drift; and a network outage may look like a period of inactivity. False negatives are equally important. A camera angle can conceal a horse against a wall, a wearable can loosen, and an algorithm trained on one breed, coat type or stable layout may perform less well elsewhere.

For these reasons, alerts should be treated as prompts for verification. The more serious the potential consequence, the more important it is to have a defined response pathway: who receives the alert, who is available to attend, how quickly they should check the horse, and what happens if the system itself is offline.

Performance and quality indicators

Buyers should request evidence rather than relying on labels such as “AI-powered”, “real time” or “vet approved”. Useful quality indicators include:

• Validation method: comparison with direct observation, video annotation or another accepted reference method.
• Population and context: breed types, ages, housing systems, number of horses, season and duration of testing.
• Performance measures: sensitivity, specificity, precision, false-alarm rate, missed-event rate and confidence intervals where available.
• Attachment safety: materials, edges, fasteners, breakaway arrangements, water resistance, impact resistance and inspection requirements.
• Environmental specification: operating temperature, humidity limits, dust and water ingress rating, cleaning compatibility and sensor placement requirements.
• Power performance: expected battery life under stated sampling and transmission settings, charging time and behaviour when the battery is low.
• Data access: export formats, retention period, timestamp accuracy, audit trail and whether raw data are available.
• Service continuity: offline recording, outage notifications, firmware support and the manufacturer’s policy if the cloud service is withdrawn.

Ingress protection ratings such as IP ratings can describe resistance to dust and water under specified test conditions, but they do not guarantee survival against biting, rubbing, pressure washing, urine, disinfectants or repeated impact. Mechanical protection and maintainability remain installation responsibilities.

Installation in a stable environment

A good installation begins with a risk assessment. Consider doors, kick zones, bedding, water, electrical supplies, fire escape routes, rodent damage, cleaning routines, lightning exposure and the possibility of a horse reaching the equipment. Avoid creating projections, loops or accessible cables. Battery compartments should be secure and any fasteners should be inspectable.

Lighting is often overlooked. A camera may function well during daylight but produce poor images when the horse stands in a backlit doorway or when infrared illumination reflects from dust, cobwebs or nearby surfaces. Test at the times when monitoring is most needed, including overnight and during routine cleaning.

Networks should be surveyed rather than assumed. Measure signal strength in every relevant stable, not just at the router. Where multiple cameras are used, calculate aggregate bandwidth and storage requirements. A local network can become congested when several devices upload high-resolution video simultaneously.

Commissioning should include a documented acceptance test. Check each sensor’s identity, time synchronisation, alert route, power-failure response, network-failure response, data retention, cleaning procedure and physical security. Repeat the test after major changes to the yard, router, stable doors or camera positions.

Cybersecurity, privacy and governance

Connected monitoring systems collect information about animals, staff, visitors, routines and sometimes private areas of a property. Commercial yards may also process client information. Security therefore matters even when the system is not controlling machinery.

Minimum controls include unique passwords, multi-factor authentication where offered, prompt firmware updates, a separate network for IoT equipment, restricted user permissions, encrypted connections, secure physical access to gateways and a documented process for removing former staff or clients. The UK government’s Code of Practice for consumer IoT security recommends principles including eliminating universal default passwords, providing a vulnerability disclosure policy and making software updates appropriately transparent. UK Code of Practice for consumer IoT security.

Manufacturers should explain what data leave the premises, where they are stored, how long they are retained, who can access them and whether they are used to train or improve algorithms. NIST’s IoT guidance emphasises clear documentation of device data collection, use, sharing and supporting services. NIST IoT device cybersecurity documentation guidance.

In a commercial setting, cameras can engage UK data-protection and employment considerations, particularly if people are visible or recorded. Operators should establish a lawful and proportionate purpose, limit access, display appropriate notices where required, set retention periods and avoid using animal monitoring as an unexplained form of staff surveillance. Specialist legal advice may be appropriate for a particular business model.

Selection by use case

Common misconceptions

“A camera means the horse has been checked.”

A camera provides an opportunity to observe, not proof that an observation occurred or that the image was interpretable. A scheduled physical inspection remains necessary, especially for welfare checks, injury assessment, equipment inspection and tasks that cannot be completed remotely.

“An alert means something is wrong.”

An alert means that a rule or algorithm detected a specified pattern. It may be caused by normal behaviour, sensor movement, a change in lighting or a communications fault. Alerts need context and a response protocol.

“More data are always better.”

More data can increase storage costs, battery use, privacy exposure and alert fatigue. A smaller number of reliable measurements linked to clear decisions is usually more useful than a large volume of poorly interpreted data.

“AI has recognised the horse’s health.”

Most systems infer activity or behaviour from indirect signals. Unless a product has been validated for a defined purpose in a defined population, it should not be presented as detecting disease or replacing veterinary assessment.

Operating a monitoring system well

Technology works best when integrated into ordinary yard management. Record the horse’s normal routine, check sensor attachment and condition, review battery status, test alerts, and log outages or unusual readings. Assign responsibility for responding to notifications. At a commercial yard, include the system in staff induction and handover procedures.

Review whether the system changes decisions in a useful way. If alerts are repeatedly ignored, the thresholds or workflow may be wrong. If staff spend time reviewing footage without a defined purpose, the system may be generating reassurance rather than information. Periodic review should ask whether the device remains safe, supported, calibrated where relevant, and appropriate to the horse’s current management.

The most defensible role for connected monitoring is as an additional layer between scheduled observations. It can improve continuity, reveal trends, support communication and provide evidence about what happened. It cannot provide continuous certainty, guarantee welfare, or remove the need for competent people to see, handle and assess horses directly.

Sources and further reading

• Department for Environment, Food and Rural Affairs: Code of practice for the welfare of horses, ponies, donkeys and their hybrids
• Department for Environment, Food and Rural Affairs: Keeping horses commercially
• Validation of triaxial accelerometers to measure the lying behaviour of adult domestic horses
• Suitability of solar-powered acceleration and global positioning devices for remote monitoring of equine welfare
• Measuring Activity-Rest Condition by Using Two Different Accelerometer Devices in Trained and Retired Horses
• Impact of the technology to monitor horse behaviour and health: a scoping review
• UK government: Code of Practice for consumer IoT security
• NIST: IoT device cybersecurity documentation guidance
• UK government: Advisory standards for the care and accommodation of animals in science

Related research

Data Quality and Limitations in Equine Sensor TechnologyContinue research →GPS Tracking Technologies in Equestrian UseContinue research →Equine Wearables and Activity Sensors ExplainedContinue research →