A machine learning-based positioning method for poultry in cage environments
Introduction
Individuals or groups of animals engage in a variety of activities. As their ranges often overlap, they generally tend to mutual avoidance rather than repelling each other. Numerous studies have demonstrated a connection between domain behavior and animal social hierarchies (Milewski et al., 2022). A dominance hierarchy develops among congregated roosters, with the highest-ranking rooster being dominant and having territorial priority (Collias et al., 1966). Dominant roosters maintain a fixed territory, whereas lower-ranking roosters retain only a small area adjacent to themselves (McBride et al., 1969). Higher-rank roosters occupy significantly more cage space and dominant areas than lower-rank roosters (Odén et al., 2004). Individuals in social species frequently form dominance relationships, in which dominant individuals have greater access to resources than subordinate individuals (Favati et al., 2014). Male domestic fowl explore new areas more quickly and maintain long-term vigilance than female domestic fowl. In contrast, abnormal walking patterns are a strong indicator of decreased welfare in broiler chickens. It is critical to understand how captive birds utilize their space to promote welfare by optimizing space quality and meeting their biological needs (Aydin, 2016).
Various technologies have been developed for indoor positionings, such as satellite-based systems (Obeidat et al., 2021), inertial navigation systems (INS), magnet-based systems, and radio frequency-based systems (Denis et al., 2019). However, many of these methods are ineffective in locating small animals. For instance, because of building exterior walls, the global positioning system (GPS), one of the most widely used satellite-based navigation systems, cannot locate indoor objects (Nirjon et al., 2014). INS can be error-prone, necessitating sophisticated filtering techniques such as the Kalman filter (Hu et al., 2020). Magnetic technology is precise but susceptible to conductive and ferromagnetic materials at low frequencies (Diaz et al., 2019). In addition, magnetic-based navigation systems typically rely on disturbances in the Earth’s magnetic field within enclosed environments, resulting from the ferromagnetic nature of metal structures within buildings (Shu et al., 2015). As the Internet of Things (IoT) technology has advanced in recent decades, radio frequency-based systems have been rapidly developed, becoming more suitable for indoor positionings, such as frequency modulation technology (Popleteev, 2017), Wi-Fi (Bagosi and Baruch, 2011), ZigBee (Bianchi et al., 2018), Bluetooth (Wang et al., 2015), radio frequency identification (RFID) (Tesoriero et al., 2010), and LoRa (Islam et al., 2019).
Among these technologies, ultra-high frequency (UHF)-RFID has been applied to agricultural positioning research because of its long identification distance, high accuracy, rapid response time, and strong anti-interference capability. UHF-RFID-based positioning methods are classified as range-based (distance measurement and angle measurement methods) and range-free (fingerprinting and non-fingerprinting methods) (Li et al., 2019). Based on the electromagnetic propagation regulation, conventional range-based methods can convert the received signal strength indicator (RSSI), time-of-flight, and phase information to the distance between the reader antenna and the target tag. The geometric characteristics infer the position of the target tag. Ma et al. (2016) increased the accuracy to 0.769 m using the phase of arrival (POA) and time of arrival (TOA) modeling. Povalac and Sebesta (2011) achieved an average absolute error of 0.14 m by improving the phase-based ranging technology.
Subsequently, methods for localization based on reference tags and fingerprinting, which are more suitable for complex environments, have been extensively investigated. Ni et al. (2003) proposed a LANDMARC system based on reference tags, which involved deploying numerous fixed-position reference tags, reading RSSI values relative to target tags, and calculating target tag positions using k-nearest neighbor (KNN). Xu et al. (2017) improved the LANDMARC system (Ni et al., 2003), achieving an average positioning error of less than 15 cm. Zhao et al. (2017) and Mo and Li (2019) proposed improved methods based on virtual reference tags and phase variations that improved localization accuracy to within 10 cm.
However, the environment in massive farms is highly complex, with issues such as temperature and humidity fluctuations, signal masking, and target tag rotation. Most methods previously discussed are based on sophisticated computational models and require stringent experimental conditions. Moreover, they lack real-time performance and robustness. Therefore, this study proposes a machine learning-based localization method to achieve accurate localization in a complex cage environment using a relatively simple experimental design. With the aid of UHF-RFID devices and our method, it is possible to precisely locate multiple chickens in metal cages. To the best of our knowledge, no other study has used the same device to achieve localization accuracy comparable to that of our method in a similar environment.
Section snippets
Experimental design
Natural mating colony cages with dimensions of 1.2 m × 1.2 m × 0.65 m were used, as illustrated in Fig. 1(A), along with an RFA915-9R40-1 UHF-RFID device manufactured by ShenZhen Fuwit Technology Co. Four UHF-RFID antennas were installed on the cage ceiling with equal spacing (0.4 m), with an antenna gain of 9 DBi, an horizontal and vertical beamwidth of 90°, and all antennas facing vertically toward the cage floor, approximately 60 cm from the floor (Fig. 1B).
If antennas were mounted outside
Features of the UHF-RFID device
The performances of our UHF-RFID device were tested, and the effects of different factors on the RSSI signal were analyzed. The experimental results are shown in Fig. 7. When the relative positions of the tag and antenna were constant, the RSSI was related to the relative axial angle (Fig. 7A). In particular, the RSSI value was the maximum when the tag’s axial direction passed through the antenna’s radiation center, and when the tag returned the same RSSI in both forward and reverse directions.
Discussion
A previous study (Liu et al., 2022) proposed a phase- and RSSI-based indoor localization method for UHF-RFID tags. The method established in this study implemented the mobile antenna position along with the received phase and RSSI profiles to locate the target tag. A valid phase profile was obtained by analyzing the received RSSI (whose strength can indicate whether the signal is stable or not). After obtaining a valid phase profile based on the received RSSI, the position of the target along
Conclusions
The characteristics of UHF-RFID devices were investigated, and the main factors that influence RSSI were analyzed. RSSI was highly correlated with relative distance, as the relative angle between antennas and the tags under the same device conditions was mainly influenced by metal and closer tags.
In this study, a method to localize caged poultry based on UHF-RFID devices and machine learning algorithms is proposed. This method did not entail the complex mathematical modeling required for
Funding
This project was funded by the National Natural Science Foundation of China (Grant No. 31902209): Hebei Province, the second phase of the modern agricultural industry technology system innovation team construction project (HBCT2018150208).
CRediT authorship contribution statement
Hao Xue: Conceptualization, Methodology, Validation, Formal analysis, Investigation, Data curation. Lihua Li: Conceptualization, Writing – review & editing, Supervision, Funding acquisition. Peng Wen: Visualization, Investigation, Writing – review & editing, Supervision. Meng Zhang: Investigation, Resources, Writing – review & editing, Supervision.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
References (31)
- et al.
Locality fixation, mobility and social organization within an unconfined population of red jungle fowl
Anim. Behav.
(1966) - et al.
The social organization and behaviour of the feral domestic fowl
Anim. Behav. Monogr.
(1969) - et al.
Male rank order, space use and female attachment in large flocks of laying hens
Appl. Anim. Behav. Sci.
(2004) - et al.
Improving location awareness in indoor spaces using RFID technology
Expert Syst. Appl.
(2010) Walking behaviour and spatial use of broiler chickens
Int. Lett. Nat Sci.
(2016)- et al.
Indoor Localization by WiFi
(2011) - et al.
RSSI-based indoor localization and identification for ZigBee wireless sensor networks in smart homes
IEEE Trans. Instrum. Meas.
(2018) - et al.
A survey on detection, tracking and identification in radio frequency-based device-free localization
Sensors
(2019) - Diaz, E.M., Ahmed, D.B., Kaiser, S., 2019. A review of indoor localization methods based on inertial sensors. Geogr....
- et al.
Personality predicts social dominance in male domestic fowl
PLoS One
(2014)
Improving the heading accuracy in indoor pedestrian navigation based on a decision tree and Kalman filter
Sensors
Lorain: Making a Case for Lora in Indoor Localization
A survey on contrastive self-supervised learning
Technologies
A cost-efficient hybrid UHF RFID and odometry-based mobile robot self-localization technique with centimeter precision
IEEE J. Radio Freq. Identif.
Review on UHF RFID localization methods
IEEE J. Radio Freq. Identif.
Cited by (4)
IoT-based agriculture management techniques for sustainable farming: A comprehensive review
2024, Computers and Electronics in AgricultureValidation of an IoT System Using UHF RFID Technology for Goose Growth Monitoring
2024, Agriculture (Switzerland)SAM for Poultry Science
2023, arXiv