An RFID Based Smart Feeder for Hummingbirds

A Review of Embedded Machine Learning Based on Hardware, Application, and Sensing Scheme

Machine learning is an expanding field with an ever-increasing role in everyday life, with its utility in the industrial, agricultural, and medical sectors being undeniable. Recently, this utility has come in the form of machine learning implementation on embedded system devices. While there have been steady advances in the performance, memory, and power consumption of embedded devices, most machine learning algorithms still have a very high power consumption and computational demand, making the implementation of embedded machine learning somewhat difficult. However, different devices can be implemented for different applications based on their overall processing power and performance. This paper presents an overview of several different implementations of machine learning on embedded systems divided by their specific device, application, specific machine learning algorithm, and sensors. We will mainly focus on NVIDIA Jetson and Raspberry Pi devices with a few different less utilized embedded computers, as well as which of these devices were more commonly used for specific applications in different fields. We will also briefly analyze the specific ML models most commonly implemented on the devices and the specific sensors that were used to gather input from the field. All of the papers included in this review were selected using Google Scholar and published papers in the IEEExplore database. The selection criterion for these papers was the usage of embedded computing systems in either a theoretical study or practical implementation of machine learning models. The papers needed to have provided either one or, preferably, all of the following results in their studies-the overall accuracy of the models on the system, the overall power consumption of the embedded machine learning system, and the inference time of their models on the embedded system. Embedded machine learning is experiencing an explosion in both scale and scope, both due to advances in system performance and machine learning models, as well as greater affordability and accessibility of both. Improvements are noted in quality, power usage, and effectiveness.

Manipulating actions: a selective two‐option device for cognitive experiments in wild animals

Advances in biologging technologies have significantly improved our ability to track individual animals' behaviour in their natural environment. Beyond observations, automation of data collection has revolutionized cognitive experiments in the wild. For example, radio-frequency identification (RFID) antennae embedded in 'puzzle box' devices have allowed for large-scale cognitive experiments where individuals tagged with passive integrated transponder (PIT) tags interact with puzzle boxes to gain a food reward, with devices logging both the identity and solving action of visitors. Here, we extended the scope of wild cognitive experiments by developing a fully automated selective two-option foraging device to specifically control which actions lead to a food reward and which remain unrewarded. Selective devices were based on a sliding-door foraging puzzle, and built using commercially available low-cost electronics. We tested it on two free-ranging PIT-tagged subpopulations of great tits Parus major as a proof of concept. We conducted a diffusion experiment where birds learned from trained demonstrators to get a food reward by sliding the door either to the left or right. We then restricted access of knowledgeable birds to their less preferred side and calculated the latency until birds produced solutions as a measure of behavioural flexibility. A total of 22 of 23 knowledgeable birds produced at least one solution on their less preferred side after being restricted, with higher-frequency solvers being faster at doing so. In addition, 18 of the 23 birds reached their solving rate from prior to the restriction on their less preferred side, with birds with stronger prior side preference taking longer to do so. We therefore introduce and successfully test a new selective two-option puzzle box, providing detailed instructions and freely available software that allows reproducibility. It extends the functionality of existing systems by allowing fine-scale manipulations of individuals' actions and opens a large range of possibilities to study cognitive processes in wild animal populations.

How feedback and feed-forward mechanisms link determinants of social dominance

In many animal societies, individuals differ consistently in their ability to win agonistic interactions, resulting in dominance hierarchies. These differences arise due to a range of factors that can influence individuals' abilities to win agonistic interactions, spanning from genetically driven traits through to individuals' recent interaction history. Yet, despite a century of study since Schjelderup-Ebbe's seminal paper on social dominance, we still lack a general understanding of how these different factors work together to determine individuals' positions in hierarchies. Here, we first outline five widely studied factors that can influence interaction outcomes: intrinsic attributes, resource value asymmetry, winner-loser effects, dyadic interaction-outcome history and third-party support. A review of the evidence shows that a variety of factors are likely important to interaction outcomes, and thereby individuals' positions in dominance hierarchies, in diverse species. We propose that such factors are unlikely to determine dominance outcomes independently, but rather form part of feedback loops whereby the outcomes of previous agonistic interactions (e.g. access to food) impact factors that might be important in subsequent interactions (e.g. body condition). We provide a conceptual framework that illustrates the multitude potential routes through which such feedbacks can occur, and how the factors that determine the outcomes of dominance interactions are highly intertwined and thus rarely act independently of one another. Further, we generalise our framework to include multi-generational feed-forward mechanisms: how interaction outcomes in one generation can influence the factors determining interaction outcomes in the next generation via a range of parental effects. This general framework describes how interaction outcomes and the factors determining them are linked within generations via feedback loops, and between generations via feed-forward mechanisms. We then highlight methodological approaches that will facilitate the study of feedback loops and dominance dynamics. Lastly, we discuss how our framework could shape future research, including: how feedbacks generate variation in the factors discussed, and how this might be studied experimentally; how the relative importance of different feedback mechanisms varies across timescales; the role of social structure in modulating the effect of feedbacks on hierarchy structure and stability; and the routes of parental influence on the dominance status of offspring. Ultimately, by considering dominance interactions as part of a dynamic feedback system that also feeds forward into subsequent generations, we will understand better the factors that structure dominance hierarchies in animal groups.

Evaluating the accuracy and biological meaning of visits to RFID-enabled bird feeders using video

Radio-frequency identification (RFID) technology has gained popularity in ornithological studies as a way to collect large quantities of data to answer specific biological questions, but few published studies report methodologies used for validating the accuracy of RFID data. Further, connections between the RFID data and the behaviors of interest in a study are not always clearly established. These methodological deficiencies may seriously impact a study's results and subsequent interpretation. We built RFID-equipped bird feeders and mounted them at three sites in Tompkins County, New York. We deployed passive integrated transponder tags on black-capped chickadees, tufted titmice, and white-breasted nuthatches and used a GoPro video camera to record the three tagged species at the feeders. We then reviewed the video to determine the accuracy of the RFID reader and understand the birds' behavior at the feeders. We found that our RFID system recorded only 34.2% of all visits by tagged birds (n = 237) and that RFID detection increased with the length of a visit. We also found that our three tagged species and two other species that visited the feeders, American goldfinch and hairy woodpecker, retrieved food in 79.5% of their visits. Chickadees, titmice, nuthatches, and woodpeckers spent, on average, 2.3 s at feeders to collect one seed per visit. In contrast, goldfinches spent an average of 9.0 s at feeders and consumed up to 30 seeds per visit. Our results demonstrate the importance of confirming detection accuracy and that video can be used to identify behavioral characteristics associated with an RFID reader's detections. This simple-yet time-intensive-method for assessing the accuracy and biological meaning of RFID data is useful for ornithological studies but can be used in research focusing on various taxa and study systems.

A Smart Sensing System of Water Quality and Intake Monitoring for Livestock and Wild Animals

This paper presents a water intake monitoring system for animal agriculture that tracks individual animal watering behavior, water quality, and water consumption. The system is deployed in an outdoor environment to reach remote areas. The proposed system integrates motion detectors, cameras, water level sensors, flow meters, Radio-Frequency Identification (RFID) systems, and water temperature sensors. The data collection and control are performed using Arduino microcontrollers with custom-designed circuit boards. The data associated with each drinking event are water consumption, water temperature, drinking duration, animal identification, and pictures. The data and pictures are automatically stored on Secure Digital (SD) cards. The prototypes are deployed in a remote grazing site located in Tucumcari, New Mexico, USA. The system can be used to perform water consumption and watering behavior studies of both domestic animals and wild animals. The current system automatically records the drinking behavior of 29 cows in a two-week duration in the remote ranch.

Use of RFID technology to characterize feeder visitations and contact network of hummingbirds in urban habitats

Despite the popular use of hummingbird feeders, there are limited studies evaluating the effects of congregation, sharing food resources and increased contact when hummingbirds visit feeders in urban landscapes. To evaluate behavioral interactions occurring at feeders, we tagged 230 individuals of two species, Anna's and Allen's Hummingbirds, with passive integrated transponder tags and recorded their visits with RFID transceivers at feeders. For detecting the presence of tagged birds, we developed an RFID equipped feeding station using a commercially available antenna and RFID transceiver. Data recorded included the number of feeder visits, time spent at the feeder, simultaneous feeder visitation by different individuals, and identifying which feeders were most commonly visited by tagged birds. For the study period (September 2016 to March 2018), 118,017 detections were recorded at seven feeding stations located at three California sites. The rate of tagged birds returning to RFID equipped feeders at least once was 61.3% (141/230 birds). Females stayed at feeders longer than males per visit. We identified primary, secondary and tertiary feeders at Sites 2 and 3, according to the frequency of visitation to them, with a mean percentage of 86.9% (SD±19.13) visits to a primary feeder for each tagged hummingbird. During spring and summer, hummingbirds visited feeders most often in morning and evening hours. Feeder visits by males overlapped in time with other males more frequently than other females. The analysis of the contact network at the feeders did not distinguish any significant differences between age or sex. Although most hummingbirds visited the feeders during the daytime, our system recorded night feeder visitations (n = 7 hummingbirds) at one site. This efficient use of RFID technology to characterize feeder visitations and contact networks of hummingbirds in urban habitats could be used in the future to elucidate behaviors, population dynamics and community structure of hummingbirds visiting feeders.