ADS 6523 Final

Artificial intelligence

intelligence displayed by machines, as opposed to natural intelligence demonstrated by humans and other animals. When a machine mimics cognitive functions typically associated with humans such as planning, learning, reasoning, problem solving, knowledge representation, perception, motion, manipulation, social intelligence, and creativity

Machine learning

a branch of artificial intelligence that uses algorithms for statistical prediction and inference and allows computer algorithms to progressively learn from senor big data sets and improve themselves accordingly, eliminating the need for a human data analyst.

Characteristics of Machine learning

• different than data mining (which gleans information from databases by finding patterns)

• ML teaches the computer how to make sense of the data (trained on a training dataset, learns to make predictions about new data)

• ML used in animal genetics research (to predict phenotypes based on genotypic information, identifying outliers in a population, and genotype imputation

• ML used in dairy production (mastitis detection from automated milking, monitor and predict lameness, estimate body weight via image analysis, monitor microbiome health)

Neural network

a computer system of a type of computer program that is designed to copy the way in which the human brain operates

Deep learning

a type of artificial intelligence that uses algorithms (sets of mathematical instructions or rules) based on the way the human brain operates

Characteristics of Neural networks

• mimic the human brain through a set of algorithms

• Main components are inputs, weights, a bias or threshold, and an output

• Output of one layer is passed in the next network layer

Characteristics of deep learning

• a neural network taht consists of 3 or more layers (each layers includes inputs and the output)

• Less dependent on human intervention and structured data to learn than machine learning (supervised learning)

• Requires more data than machine learning to improve accuracy

Difference between preedictive and reactive paradigms

Reactive (has already happened, less options), Proactive (what is happening), Predictive (what may happen, looking ahead)

Data analytics maturity curve

Hindsight (descriptive - what happened), insight (predictive - what will happen), foresight (prescriptive - what can happen)

Current technologies in animal agriculture

• robotic milkers

• wearable senors

• radio frequency ID ear tags

• Feed sensors

  • (all for individual animals and minimal human intervention)

Advanced technology collection and analysis process

1. sensors (capture)

2. Big data (processes)

3. Machine learning (analyzes)

4. Algorithm (notifies)

Benefits of advanced technology collection

• facilitate collection and analysis of more diverse datasets for improved accuracy

• consider variables such as genetics, environment and management priorities

• develop relevant and contextually optimal solutions

• reduce costs, increase production, enhance efficiencies, and create optimal formulations

• provide evidence-based data and data-driven solutions

Key technical challenges in machine learning

1. computational complexity (challenges in algorithm selection, higher storage requirements and energy use)

2. Concept drifts (required adaptation to changes in data distributions within a concept)

3. Discrepancies in different situations (retraining of system needed when in new environment or with new animals)

Potential concerns with machine learning

1. insufficiently modeled algorithms (can have unintended and adverse effects)

2. Data outliers (example: 100 year storm data outliers)

3. Inherent biases in data (can effect predictive models)

technology convergence

• blurring of the boundaries between disjoint areas of science, technology, markets, or industries. Synonyms are technology fusion, merging, cross fertilization, hybridization, and overlap

• Where 2 or more independent technologies integrate and form a new outcome

• Merging, blending, integrating, and/or transforming technologies leads to a completely new converged technology

• Converged technologies may replace or render obsolete single-function technologies

• Progression of evoluation of technology

• 3 characteristics:

  • can execute multiple functions to serve blended purpose

  • can collect and use data in various formats and employ machine learning techniques to deliver enhanced user experience

  • Connected to a network directly and/or are interconnected with other devices to offer ubiquitous access to users

IOT (internet of things) is actually a

covergence of technologies

Upper layers are made of

users interact more with these layers

Tech convergence takes science and leads to

solutions

Technology convergence includes fusing information from multiple sensors

• data fusion is based on the independent observation data of multiple sensors. More effective information can be obtained through composite application and algorithm design, to eliminate the limitation that a single sensor can only obtain partial information of detection target

Features of multi-sensor fusion technology

• strong environmental adaptability

• wide perception dimension

• high measurement accuracy

• fast processing speed

• complex structure

Smart farming vs. precision ag

both precision ag and smart farming refer to the use of modern technologies such as the Internet of Things (IOT), drones, robotics and Artificial intelligence (AI) in the control and management of farms in order to improve productivity adn yield, while reducing input, land, and labor requirements

• PA is distinct from SF in that it specifically refers to IOT-based approaches aiming at improving the efficiency of input use through providing farmers with tools that increase the granularity of decision making

Precision livestock farming (PLF)

management of individual animals by continous automated real-time monitoring of livestock production/reproduction, health, welfare, and environmental impact

Complex, individual, and time variant (CIT) system

a complex, individually different and time variant system; applies to all living organisms, including animals

Bioresponse

the biological response of an animal when exposed to external stimuli

Animal variable

any parameter related to the behavioral or physiological state of the animal (like weight, activity, feed intake,; typically acquired in the field)

Feature variable

variable that may be calculated based on measured animal variables and that describes the bio-response of interest

Target variable

variable that is derived from feature variables and that relates to the final objective of the PLF method; used as foundation for making management decision

Gold standard

reliable and/or generally accepted method of measuring or observing target variables; may be expensive, complex, and difficult to asses, but as necessary to verify that a PLF method provides reliable outputs

Considerations specific to animals to account for in PLF

• animal observation and monitoring is more challenging (animals at times display complex behaviors that may be difficult to observe and intepret)

• animals may move or not cooperate with the manager (animal movement or lack of coopration with handlers may make implemenation of automated actions more difficult)

• animal needs beyond survival are linked with certain behaviors (animals must be able to show certain behaviors and maintain a certain perceived quality of life or welfare)

Conditions that must be met to achieve favorable monitoring and control of PLF livestock production processes

• continuous measurement of animal variables (the information must be analyzed continously)

• availability of reliable predictions (every moment a reliable expectations must be available on how animal variables will fluctuate or animals will respond to environmental changes)

• analyzing algorithm that integrates predictions and measurements (online information is used together in mathematical equations to monitor or manage animals automatically)

Realistic use of data via PLF

PLF technoligies installed in livestock houses generate a huge amount of data. Data transmission takes time, energy, and money. Sending data wirelessly involves energy dan costs. Development of real-time algorithms that calculate information from the data at the lowest possible level enable transmission of relevant information rather than data. Algorithms are needed that can calculate relevant information from the data on or close to the individual animal

3 B’s of precision livestock farming techniques

1. Biometrics sensors

   1. monitor behavioral and physiological parameters of livestock allowing managers to evaluate an animal’s health and welfare over time

   2. Can be invasive or non-invasive

2. Big data analytics

   1. the acquisition (often from biometric sensors) and analysis of large, complex datasets

   2. Used to inform management decisions

3. Blockchain

   1. decentralized or distributed encrypted transactions ledger technology

What PLF technologies can facilitate

1. individualized animal care (provide real time alerts to deliver individualized (and even automated animal care)

2. Novel phenotype measurements (measure phenotypes that would be difficult to measure without digital technologies)

3. In-depth analysis of animals within their environment (help genetic improvement of adapted animals for different environments and development of enhanced animal facilities and equipment)

4 phases of PLF operational processes

1. observe (intelligent senors, online monitoring)

2. interpret (predicitive models, system simulations)

3. decide (decision support systems, artificial intelligence)

4. act (automated actuation, integrated operations)

inner cycle of 4 phases of PLF operational processes

represents the present state of the art industry, with manual actions and monitoring, and experience based interpretation and decision making

outer cycle of 4 phases of PLF operational processes

illustrates how the introduction of PLF may influence the different phases of the cycle

Internet of Things (IOT)

a network of objects that are connected wirelessly using sensors, and can transmit information to each other, or a wider network, without human intervention; connected objects include humans, animals, plants, and infrastructure (ex: equipment, buildings, etc.)

Sensor

a device that measures or detects a biological, chemical, physical, or mechanical property, or a combination of these properties, and records and collects the data for intepretation by a human or a machine

Acoustic sensor

measures sound, microphones convert sounds into electrical signals

Chemical sensors

measure (part of) the chemical composition of a liquid or gas environment, must recognize a chemical compound, must give information on chemical compound concentration

electrical sensor

measures electrical current and conductivity

pressure sensor

sensor consists of a thin, flexible membrance that can deform and covers a reference cavity sealed at low vacuum pressure

flow sensor

mass flow meter measures flow rate of a fluid through a tube, mass flow rate divided by fluid density to calculate volume per unit of time, volumetric flow rate determined based on fluid conductivity

mechanical sensor

measures poistion, angle or accerelation (change in velocity over time), accelerometer measures accerelation, gyroscope measures angle, inclinometer (tilt sensor, slope sensor , clinometer) measures object angle using force of gravity

temperature sensor

measures climate conditions of animal’s environment or animal body temperatures, measued mechanically (via thermometers or bimetallic strips), electrically (via thermistors, themocouples, resistance thermometers, or silicon bandgap sensors) or based on infrared radiation (via thermographic cameras)

location sensors

measured by global positioning system (GPS), radio frequency identification (RFID), radio tracking, bluetooth devices, Global System for Mobile communication (GSM) devices, or wireless local area network (WIFI)

optical, imaging, and light sensors

measured in Lumen (total amount of viable light from a light source) or Lux (total amount of light that falls on a particular surface), light sensors types include photoresistors, photodiodes, and phototransistors

Cameras sensors

detect visible light and form an image, based on optical sensors, involve vision and image recognition systems, digital cameras use software to interpret grid electrical charges generated by light conditions and determine dark and light image areas along with colors

Sensors can be

handled, attached to an animal, mounted or placed in an animal’s environment

computer vision

field that aims to describe the world through images by interpreting, reconstructing, and extracting properties from images, such as shapes, textures, densitites, and distances; also known as machine vision systems, viisual image systems, or image systems

digital image processing

capturing and translating a visual signal into a digital image; involves the conception, design, development, and enhancement of digital imaging algorithms and programs

Digital image analysis

process of extracting meaningful information from an image

Images on visible light spectrum

capture electromagnetic waves withing the visible light spectra to generate digital images, measured by standard digital cameras and/or survelliance cameras

Infrared Radiation (IR)

wavelength longer than visible light, divided into near-infreread (NIR), mid-infrared (MIR), and far-infrared (FIR), based on specific wavelength thresholds, can be used for 3D imaging spectroscopy, night vision, and thermal imaging (thermography)

3D Imaging

can be used for measurement of animal volume, surface, and gait

• passive 3D imaging

• stereo imaging uses 2 or more cameras to calculate the distance of a point to the cameras

• structure from motion creates a disparity map between the images from a single moving camera where each image captured by the camera where each image captured by the camera is used as the distance between “cameras”. Objects of interest must be motionless

• Hybrid 3D imaging

• combine RGB (intensities of red, green, and blue) sensors with depth sensors

• Active stereoscopy combines stereo imaging from multiple cameras with structured light to improve depth estimation

• Active 3D imaging

• structured light uses active emission of knwon light patterns where the illuminated surface shows structural distortions in patterns

• Time of flight (TOF) and Light Detection and Ranging (LIDAR) cameras are based on signal modulation and ranging; measure distance between sensor and target object by detecting time difference from signal emitted by transmitter, reflected on target object, and captured back by receiver

Other types of imaging

hyperspectral imaging, ultrasound, x-ray, computed tomography (CT), magnetic resonance (MR), and satellite images

Body composition imaging requires

a trained operator

Livestock imaging examples

• ultrasound body composition scan of a bull

• thermal image of heifer using a smart phone app

• satellite image of heifer locations over time within a pasture

Principles of time of flight: 3D cameras (depth sensors)

this technique measures the distance between the senor and a target obejct by detecting the time difference from the signal emitted by a transmitter, reflected on a target object, and captured back by a receiver

Internet of services (IoS)

a collaborrative business ecosystem or global market where services from diverse providers (third parties) are offered, discovered, and consumed in shared use; a future internet that detects and uses contextual information to seamlessly adapt to an unpredicted scenario, allowing the ad hoc configuration of new information and communication technology business models

Internet of things vs. Internet of services

differences between them

Internet of Services with human interface

at least one service actor (provider or consumer) is connected with an H-M (human-machine) interface

Wearable internet of things (W-IoT)

integration of wearable devices, sensors, and communication technology in which sensors and wireless communication technology are embedded into wearable devices to collect big data for analysis and evaluation of the status of the objects monitored; an extension of Iot Technology

Features of wearable internet of things (W-IoT) technology advantageous to animal monitoring

• small size, light weight, low energy consumption, convenient integration, strong adaptability, flexibility, biocompatability

• Information perception not limited to environmental indicators but also animal physiological and behavioral status

• Can use bioenergy and behavioral movement to realize self-power supply

• Wearable microsystems attached to animal bodies can achieve precise monitoring recording, and prediction

• Good maintainability and expansibility; can quickly arrange, adjust, and collect big data at almost any time, place, and environmental condition

• Can facilitate intelligent control, remote management, and precise decision-making

What Precision Livestock Farming (PLF) can facilitate in horses

• each horse has unique phsyiology and behavior

• ability to monitor horse vital signs over long periods and during exercise can lead to the following benefits:

  • better performances - monitoring to tailor and fine tune training to meet individual needs ot the horse

  • reduced injury risks: monitoring to tailor activities, risk indentification and mitigation or avoidance

  • improved welfare: early disease detection, identification of abnormalities in daily behavioral patterns

Application of sensors and automated monitoring in horses: physical measures

• saddle pressure

  • good saddle fit to disperse pressure and enhance performance

  • poor fit may result in back pain, injuries, poor attitude to work, and reduced performance

  • May be useful to measure rider position in addition to horse measures

• Rein tension

  • quantification of applied concentrated pressure

  • Tension consistency

  • Speed of tension application and release

Application of sensors and automated monitoring in horses: physiological measures

• heart rate:

  • cardiovascular function, and indication of mental stress

• Temperature

  • indicator of physiologic status, welfare, and stress response

  • Rectal temperature

  • Central venous temperature as a stress indicator

  • Gastrointerstinal temperature

  • Thermal comfort of blanket-wearing horses

• Respiration rate

  • effects of heat, exercise, and stress on thermoregulation and breathing

Application of sensors and automated monitoring in horses: behavioral measures

• activity

  • patterns such as lying behavior (position, frequency, and duration)

• Tail movements

  • more frequent and shorter movements may indicate foaling approaching

• Head-neck position

  • angle with the vertical

  • ability to remain steady

• Location

  • distance traveled

  • speed traveled

• Gait analysis

  • lameness detection

• Grazing

  • jaw movements

Precision livestock farming (PLF) sensors used for horses

Pressure sensor and heart rate sensor

Precision livestock farming (PLF) sensors used for horses

temparature sensor, resistance, accelerometerS

Precision livestock farming (PLF) sensors used for horses

positioning and vision

Horse foraging behavior detection

combination of sound recognition and articial intelligence technologies

• can detect and distinguish chews, bites, and noise events

• provides information on bite rate, bite mass, and grazing time

• Can help determine management strategies that optimize intake

• Provides data for modeling foraging behavior to predict pasture use and animal performance

Horse gait detection via human wearables

• schematic of a system using a smart watch and smart phone to recognize horse gaits

• Illustratres the concept that information about animals can be collected from measurements on humans engaged in the human-animal interface

Example of alert technology in animal applications: Foaling alert system

• components

  • main unit/repeater

  • antenna/cable

  • pocket pager

  • transmitter with protective pouch to attach to mare’s halter

• Features

  • range up to 1+ mile

  • future version to include optional telephone wireless dialing component

What PLF precision livestock farming technologies help address in companion animals

increasing trends on monitoring pet lifestyle and health and focus on communication between pet and owner

• Concerns in the pet sector include:

• Alone or in a group - socialization needs of dogs, antagonistic interactions in multi-cat households

• Indoors versus outdoors - concerns about traffic accidents involving pets and behaviors when owner not home

• Breeding - appropriate breeding strategies considering health concerns, behavior traits

• Therapy and assistance animals - monitoring welfare of animals trained to perform specific tasks

Application of senors and automated monitoring in pets: physiological measures

• heart rate

  • cardiovascular function and indication of mental stress

  • measure of activity

• Temperature

  • indicator of health and disease incidence

  • rectal temperature

  • non-contact infrared camera detection

• Respiration rate

  • important health indicator

  • effects of heat, exercise, and stress on thermoregulation and breathing

• Glucose

  • diabetes monitoring using sensor implanted under skin

Application of senors and automated monitoring in pets: behavioral measures

• activity

  • accelerometers used to measure activity levels

• Specific behaviors and emotions

  • distinguish between active behaviors such as walking, trotting, galloping, eating, head shaking, sniffing, and jumping

• Location and use of space

  • animal whereabouts and home range

  • finding runaways

  • theft deterrent

Application of senors and automated monitoring in pets: sound measures

Vocalizations (barking in dogs)

• training to deter barking

• distinguish emotional states

Overview of Precision livestock farming (PLF) sensors used for companion animals

camera

heart rate sensor

temperature sensor

Overview of Precision livestock farming (PLF) sensors used for companion animals

accelerometer

respiration sensor

weighing sensor

Overview of Precision livestock farming (PLF) sensors used for companion animals

GPS

Chemical sensor

Vibration sensor

Pet wearables

• fitness tracker features are bonus features to the core purposes of GPS tracking devices which is to locate a pet in the event it gets separated from its caregiver

• activity tracking incentives should encourage activity and friendly competition but not “overwalking” a dog

• Activity tracking information can be shared with a pet’s veterinarian to give actual data instead of estimates