Livestock farming (LF) plays a critical role in the global economy due to the rapidly growing pressure on food supply (Gray et al., 2017; Barbedo and Koenigkan, 2018; Alanezi et al., 2022a; Alanezi et al., 2022b). In reaction to this global trend, the productivity (including input and capacity efficiencies) of LF industries must scale up proportionally (Mudziwepasi and Scott, 2014; Sharma and Koundal, 2018; Friha et al., 2021). Low-cost automated wireless sensor network (WSN) technologies have served as viable solutions to tackle this production capacity challenge (Kakamoukas et al., 2019; Xiang et al., 2019; Boursianis et al., 2020; Long et al., 2020; Loukatos and Arvanitis, 2021; Alanezi et al., 2022a; Loukatos et al., 2022). Moreover, the integration of embedded Ambient Intelligence (AmI) services into Internet of Things (IoT) devices for smart sensor nodes (SNs) have led to the effective allocation, dynamic management, and distributed control of scarce energy resources (Barriuso et al., 2018; Mistry et al., 2022; Symeonaki et al., 2022). Conservation and optimal utilization of scarce energy resources is central to prolonging the lifetime of specialized WSN applications deployed for monitoring livestock’s vital stimuli and tracking their locations (Barbedo and Koenigkan, 2018; Sharma and Koundal, 2018). One of the common energy conservation techniques is the adoption of low-power portable, wearable and autonomous SNs for sensing, gathering and processing of livestock stimulus (Barriuso et al., 2018; Casas et al., 2021; Alanezi et al., 2022b). However, unfavorable atmospheric conditions coupled with the unpredictable roaming pattern of livestock makes tracking very challenging, especially for applications deployed solely based on wearable SNs. These challenges include rapid energy depletion, high cost of continuous real-time monitoring, frequent coverage loss in large network field, short network lifespan, and monotony of manual intervention in system’s operation.

These challenges have spurred series of research efforts on the viability of adopting Unmanned Aerial Vehicles (UAVs) to tackle routing issues related to monitoring livestock stimulus (Mudziwepasi and Scott, 2014; Pajares, 2015; Gray et al., 2017; Sharma and Koundal, 2018; Freed et al., 2021; Mukhamediev et al., 2021; Petrova and Petrov, 2021; Mistry et al., 2022). The key technical benefits of incorporating UAV into LF surveillance solutions are design flexibility, cost-effectiveness, and coverage effectiveness (Chamoso et al., 2014; Chabot and Bird, 2015; Barbedo and Koenigkan, 2018; Rivas et al., 2018; Afrianto et al., 2020; Behjati et al., 2021). In such LF surveillance applications, UAVs serve as 1.) sinks for gathering high-priority stimulus from wearable SNs and 2.) flying mobile terminals for direct observation of the livestock (Xu et al., 2016; Yue et al., 2018). Therefore, the large continuous production of LF products necessitates the fusion of WSN with UAV innovations to tackle the rising global food supply pressure. Researchers have consequently deployed several integrated UAV-WSN technologies to boost the production capacity of LF industries, where UAVs serve as mobile sinks (MS) gathering livestock stimulus, and specialized Embedded Intelligence (EI) algorithms are utilized for decision-driven predictive data analytics (Yue et al., 2018; Boursianis et al., 2020). Integrated UAV-WSN solutions are therefore highly effective in improving the end-to-end LF food supply chain process (Xu et al., 2016; Alanezi et al., 2022b; Mistry et al., 2022). Most especially, distributed unsupervised livestock surveillance solutions mitigate the monotony, difficulty, and tediousness of manual inspection of large livestock farms. However, the issues of short network lifespan and frequent coverage loss still subsist.

Researchers have proposed a wide range of generic (Sivakumar et al., 1999; Heinzelman et al., 2000; Heinzelman et al., 2002; Bandyopadhyay and Coyle, 2003; Ding et al., 2005; Huang and Wu, 2005; Lotfinezhad and Liang, 2005; Neethirajan, 2017) and specific (Rahmatizadeh et al., 2014; Asmaa et al., 2016; Nitesh et al., 2017; Yue et al., 2018; Zhang et al., 2018; Behera et al., 2019; Huang et al., 2020; Gnanasekera et al., 2021; Lin et al., 2021; Hu et al., 2022; Quoc et al., 2022) routing techniques to address these issues. Among the proffered techniques, cluster-based routing (CBR) strategies have relatively exhibited better network resilience, connectivity, adaptability, reliability, energy-efficiency, and computational simplicity (Salami et al., 2009; Salami et al., 2010; Astakhova, 2020; Alanezi et al., 2022b).

Effective design of distributed unsupervised livestock monitoring applications using CBR strategy requires flexible, scalable, low-cost, and energy-aware routing architectures. This is usually implemented by dynamically, logically, and systematically arranging smart SNs fastened to livestock herds into sets named herd clusters (HCs) in this research work, for easily achieving network scalability, adapting to dynamic topology and conserving scarce energy resources. For each HC, the network elects HC leads (HCLs) based on their relative distance cost with respect to the MS, residual energy, and shortest-path accessibility for data transmission. After the cluster configuration phase, HC members (HCMs) forward sensed livestock stimulus to their respective HCLs using single-hop or multi-hop transmission. After the intra-cluster data transmission phase, HCLs forward cumulated data to the hovering MS for subsequent relaying to the base station (BS) (Gnanasekera et al., 2021; Yaxley et al., 2021; Alanezi et al., 2022b; Hu et al., 2022; Li et al., 2022).

Centered on this herd CBR methodology, this article proffers a recoverable UAV-based energy-efficient reconfigurable (RUBER) routing algorithm for smart wireless livestock sensor networks. This proposed RUBER scheme is an improvement on the previous UBER scheme developed by the same authors. One of the major improvements and technical differences of RUBER is that it specifically solves sensor node and route failure issues which are significant challenges faced by the UBER scheme. Another technical difference lies in the unique features, techniques and operational mechanisms introduced and utilized by these CBR schemes. RUBER relies on an efficient fault detection and recycling technique, dynamic recovery mechanism, and robust route maintenance technique while UBER relies on UAV-assisted coverage recovery mechanism, dynamic residual energy thresholding strategy, and robust cluster-to-UAV link formation. These operational mechanisms and techniques introduced for RUBER are specialized enhancements, functional improvements and algorithmic extensions over the operational features of UBER. MATLAB and OMNET++ simulation experiments were employed for performance assessment and comparative analysis of RUBER by benchmarking its performance with UBER and HYBRID. Simulation experiments demonstrated the efficiency and relative competitiveness of the RUBER scheme against the benchmarked schemes.

The rest of this article is arranged in the following way: Section 2 reviews the pertinent fault-tolerant CBR strategies for MS-based applications. Section 3 technically describes the mechanism of operation of the proposed RUBER protocol, while Section 4 delivers the ensuing simulation results together with buttressing explanations. Section 5 provides the conclusion to this article.

Conventional CBR strategies are usually categorized as generic based on their network architecture, routing mode, and compatibility with diverse WSN applications. However, the common limitations of generic CBR schemes are cluster re-configuration/re-sizing issues for large-scale networks, time-complexity tradeoff, rapid energy depletion due to long-range single-hop transmissions, and frequent HCL failures (Sivakumar et al., 1999; Heinzelman et al., 2000; Heinzelman et al., 2002; Bandyopadhyay and Coyle, 2003; Ding et al., 2005; Huang and Wu, 2005; Lotfinezhad and Liang, 2005; Salami et al., 2009; Salami et al., 2010; Bello-Salau et al., 2011a; Bello-Salau et al., 2011b; Salami et al., 2011; Neethirajan, 2017).

UBER was proffered as an energy-efficient, coverage-aware and reconfigurable CBR technique for smart wireless livestock sensor networks (Alanezi et al., 2022b). This CBR technique relies on UAV-assisted coverage maintenance mechanism, adaptive cluster-to-UAV link formation, and linearized energy thresholding pattern for reduced energy consumption. Clustering costs and proximity to aerial MS are part of the parameters for HCL election while multi-hop data forwarding chain is dynamically constructed for end-to-end communication in the network (Alanezi et al., 2022b). The drawback of this CBR technique is the non-availability of repair and recovery strategy to cater for the detected SN and route failures.

FCGW was proposed as a fault-tolerant CBR scheme that relies on multi-path routing, symmetric link formation, Gaussian network architecture, and square-shaped cluster formation (Quoc et al., 2022). The LF field is divided into virtual square grids (or clusters) while Gaussian integers are used for electing the HCLs and MS are employed for distributed load balancing and network density re-adjustment. Shortest inter-cluster distance costs are computed for end-to-end data transmission with the aid of multiple path formation in the Gaussian network (Quoc et al., 2022). The limitations of this CBR scheme are the broadcast overhead issues and large delay associated with establishing long-range transmissions in the Gaussian network architecture.

HYBRID was proffered as a CBR strategy for network lifetime enhancement by engaging SNs deployed in unfavorable LF environments where energy consumption is managed by systematically partitioning the LF field into logical HCs and situating the BS at the edge of the network (Behera et al., 2019). Residual energy of SNs is the basis for constructing HCs. This means that SNs with relatively more residual energy have higher probability of being chosen as HCLs. Distance costs between HCs are taken into consideration for multi-hop transmission of livestock stimulus in the LF network (Behera et al., 2019). The drawback of this strategy is the processing costs of operating varying transmission modes and energy levels.

MS-DVCR was offered as a geographic routing scheme for minimizing energy consumption by employing MS-based goal-oriented virtual coordinate technique (Rahmatizadeh et al., 2014). The function of the virtual coordinate technique is to minimize the number of network update occurrences as aggregated data is relayed to the BS from the MS. This routing scheme provides an effective and alternative means for achieving MS localization without the need of taking actual distance measurements on the LF field (Rahmatizadeh et al., 2014). The limitation of this routing scheme is the execution overhead associated with virtual coordinate information broadcast, maintenance, and management.

AODV-SMS (ABC-PSO) was proposed as a fault-tolerant CBR technique for route optimization in mobile WSN using artificial bee colony-particle swarm algorithm (Yue et al., 2018). This CBR scheme incorporates path coding, collaborative updating, swarm evolution, and particle selection for the recovery and optimal formation of alternate routes. The strengths of this scheme are reliable data transmission route, effective route recovery ability, and relatively more accurate computations of network parameter(s) optimization (Yue et al., 2018). The major limitations of this strategy are the computational cost for running the optimization routine and the resulting associated energy tax of this algorithm.

VBRP was put forward as a delay-efficient coverage-aware CBR algorithm for optimizing MS trajectory in critical applications (Nitesh et al., 2017). This CBR algorithm relies on local orphaned SN recovery strategy, single-hop communication, and utilizes Voronoi vertices for computing rendezvous points which aids in MS path length reduction. The strengths of this algorithm are extended network lifespan, relatively optimal MS path length, reduced average waiting duration, and better network adaptability (Nitesh et al., 2017). The limitations of this algorithm are poor energy balancing and coverage loss issues due to topology variations.

RRM-WLDCNN was proffered as a resilient routing scheme for high-reliability monitoring applications (Huang et al., 2020). This routing strategy relies on graph labelling, subgraph extraction, and path prediction technique using deep-learning (kernel and dual convolutional neural network) approach. The benefits of this scheme are lower data transmission cost, improved path reliability, enhanced network lifetime, and better energy consumption performance (Huang et al., 2020). The drawbacks of this scheme are interference issues, processing delays, and lack of flexibility and adaptability for heterogeneous and dynamic large-scale WSNs.

Table 1 provides the technical summary of the reviewed CBR schemes.

This section discusses the integrated network architecture, operational assumptions, cluster setup process, routine sensing and forwarding operations, failure detection and recycling procedure, recovery mechanism, and route maintenance strategy.

This section is a technical discussion on RUBER’s performance indices, simulation specifications, time complexity, performance assessment with respect to different failure rates, and performance analysis of RUBER in comparison with UBER and HYBRID.

This paper addresses sensor node and route failure issues for smart wireless livestock sensor networks by developing a recoverable UAV-based energy-efficient reconfigurable routing (RUBER) scheme. RUBER scheme relies on an efficient fault detection and recycling technique, dynamic recovery mechanism, and robust route maintenance technique for the running of an integrated heterogenous network. Performance-based experiments were conducted with MATLAB and OMNET++ simulations. Performance of RUBER was analyzed under low, medium and high failure rate conditions. Simulation results demonstrated that running a low-failure rate network yielded significant performance improvements with respect to failure detection ratio, failure recovery ratio, load balancing ratio, and packet delivery ratio. Performance comparison of RUBER was conducted against UBER and HYBRID schemes. With respect to best-case values, RUBER exhibited marked performance gains in terms of average routing delay, energy dissipation ratio, routing overhead, fault tolerance index, nodal failure recovered, route failure recovered, and cluster survival ratio. Results obtained revealed the relative competitiveness of the RUBER scheme against the benchmarked schemes.

Future research efforts in this project will focus on the development and deployment of an integrated UAV-WSN testbed for livestock monitoring applications distributed for experimental use in the campus of University of Hafr Al Batin. This will facilitate systematic performance evaluation and comparative assessment of actual field measurements with the obtained simulation results. Furthermore, the identified time complexity and processing cost issues of RUBER will be addressed in an upcoming research article by designing a time-aware cost-effective CBR scheme.