Boa Fumigator: An Intelligent Robotic Approach for Mosquito Control

Abstract

The mosquitoe population is reaching critical levels globally, posing significant threats to public health and ecosystems due to their role as vectors for diseases. This paper presents the development of a mobile robotic platform named Boa Fumigator with autonomous fumigation and prioritized path planning capabilities in urban landscapes. The robot’s locomotion is based on a differential drive, facilitating easier maneuverability on semi-automated planar surfaces in landscaping infrastructure. The robot’s fumigator payload consists of a spray gun and a chemical tank, which can pan and fumigate up to 4.5 m from the ground. The system incorporates a wireless charging mechanism to allow for the autonomous charging of the mosquito catchers. A genetic algorithm fused with an A*-based prioritized path planning algorithm is developed for efficient navigation and charging of mosquito catchers. The algorithm, designed for maximizing charging efficiency, considers the initial charge percentage of mosquito catchers and the time required for fumigation to determine the optimal path for charging and fumigation. The experiment results show that the path planning algorithm can generate an optimized path for charging and fumigating multiple mosquito catchers based on their initial charge percentage. This paper concludes by summarizing the key findings and highlighting the significance of the fumigation robot in landscaping applications.

1. Introduction

Mosquitoes are an integral part of the natural ecosystem. Factors such as climate change, increased vegetation and habitat, unchecked still water sources, and urbanization—including infrastructure development and landscaping in cities—have significantly increased mosquito populations. These mosquitoes bite and transmit serious diseases such as Dengue, Zika, and Malaria [1]. According to the 2023 World Health Organization report on the global dengue situation, 171,991 cases and 753 deaths were reported in Africa, while 1,895,122 cases and 2049 deaths were confirmed across North and South America combined, all attributed to Dengue [2]. Fumigation safeguards urban landscapes by controlling pests like mosquitoes, thus preventing the spread of diseases. Apart from spraying foliage, soil fumigation is another method where chemical fumes are applied to the soil to control pests, bacteria, and weeds. However, manual fumigation poses significant health risks to workers who inhale toxic smoke over prolonged periods. This exposure can lead to symptoms such as pesticide poisoning, peripheral and central neurological disturbances, or swallowing disorders [3]. To mitigate these health risks, deploying autonomous landscape fumigation robots is an effective alternative. These robots can replace human workers, ensuring safer and more efficient fumigation without the associated health hazards.

Recent advancements in the development of mosquito catchers (MCs) and fumigation techniques have been driven by the integration of machine learning, internet of things (IoT), optimized path planning, and wireless charging technologies. These innovations include fast charging solutions and reduced waiting times, supported by optimized path planning algorithms for greater efficiency. Fengjuan Tian et al. conducted a comprehensive investigation into mosquitoes and mosquito-borne viruses in Malipo and Funing counties along the Sino-Vietnam border. Their research involved the collection and isolation of 10,800 mosquitoes, identifying 29 species and 62 viral isolates across 11 categories [4]. Archana Semwal et al. presented an artificial intelligence (AI)-based mosquito surveillance and population mapping system, using the “Dragonfly” robot integrated with the you only look once (YOLO) v4 deep neural network. This system demonstrated a confidence level of 88% in detecting and classifying mosquitoes during offline tests, with an 82% confidence level in real-time field trials [5]. Mgeni M. Tambwe et al. explored alternative methods (odor lures, mosquito-electrocuting traps, and Biogents Sentinel traps) to human landing catches for evaluating the effectiveness of transfluthrin emanators against Aedes aegypti mosquitoes [6]. Meanwhile, Himanshu Sharma et al. introduced a smart agriculture monitoring system that utilized solar energy harvesting wireless sensor networks. This innovation aimed to ideally extend the network lifetime and enhance throughput for various applications [7,8]. Dingzhu Xue et al. focused on IoT-based wireless sensor networks within smart agriculture. They improved the low-energy adaptive clustering hierarchy protocol to boost node efficiency and proposed an enhanced distance vector-hop positioning algorithm [9,10,11]. Sidharth Jeyabal et al. developed a robot equipped with depth cameras, light detection and ranging (LiDAR), and inertial measurement units with advanced sensor fusion technology to accurately identify and localize the mosquito hotspots for precision fumigation to reduce chemical usage while spraying. The proposed targeted fumigation system in their study achieved 81% hotspot detection precision and observed a 62.5% increase in efficiency with reduced chemical usage [12]. The literature demonstrates the effectiveness of IoT in monitoring and managing resources in diverse applications, from agriculture to mosquito surveillance. Based on this, IoT technology is used in the current study to monitor the charge levels of MCs, enabling smart energy management and fully autonomous operation.

Optimized path planning also plays a critical role in autonomous navigation, which enhances the efficiency of reaching charging stations by prioritizing routes and minimizing travel time. Various path planning strategies exist for autonomous navigation, including cell decomposition, the roadmap approach, and the artificial potential field method [13,14]. Research in [15] demonstrated that Dijkstra’s algorithm is considered a benchmark solution, while constrained particle swarm optimization outperforms other meta-heuristic approaches in unknown environments. Yuanhao Li et al. introduced the predicting–scheduling–tracking charging scheme, which uses long short-term memory for node location prediction, a scheduling algorithm for target selection, and a Kalman filter-based tracking algorithm for distance control during charging [16,17]. Wenyu Ouyang et al. proposed an integrated data charging system strategy based on matroid theory, optimizing scheduling by considering the importance, deadlines, and penalty values to minimize data loss. Their approach includes a greedy algorithm that classifies tasks into early and delayed ones with a charging sequence adjustment method optimizing the mobile charger’s trajectory for efficiency [18]. Jihoon Lee et al. proposed a five-step, multi-level path planning system for indoor search and rescue operations that handles multi-goal path planning, maintains clearance conditions, and adapts to dynamic environmental changes [13]. Meiyan Zhang et al. enhanced Q-learning-based coverage path planning for mobile robots by incorporating new reward functions inspired by the predator–prey model [19,20]. Longda Gao et al. introduced a complete coverage path planning algorithm incorporating energy compensation and obstacle vectorization [21]. Anirudh Krishna Lakshmanana et al. presented a reinforcement learning-based complete coverage path planning model for the hTetro reconfigurable robot, achieving lower costs and faster path generation compared to traditional models in arbitrary environments [22]. A* serves as the foundational algorithm for all robot path planning methods and is utilized extensively in the literature above. Including a genetic algorithm (GA) with the A* algorithm in this study enhances path planning by dynamically prioritizing charging multiple MCs. This allows for real-time adjustments, optimizing travel time and distance while maintaining MC charge above 30%.

Implementing wireless charging into the MC ensures that the MC system remains fully enclosed and protected from environmental factors such as rain and dust. Ying Dong et al. optimized wireless sensor network charging by considering node energy, location, and communication modes. Simulations show improved efficiency, reduced frequency, and balanced energy consumption [23,24]. Some related work has recently been carried out on developing robots equipped with wireless charging capabilities. Ivan Cortes et al. developed a system that employs eight sensing coils to detect lateral misalignment between a mobile robot and a moving wireless charger. A dual-loop control system with a linear quadratic regulator maintains alignment within 2 cm during experiments with a 5 W charger moving at 0.145 m/s. [25]. Chi Lin et al. introduced a game-theoretical approach for wireless rechargeable sensor networks using wireless charging vehicles [26]. Zhen Wei et al. developed a method to minimize dead sensor nodes and maximize energy utilization in a wireless charging environment with limited energy resources [27]. Sheng Zhang et al. optimized wireless charging with pre-planned itineraries for rechargeable devices [28]. The literature shows wireless charging offers efficient, hands-free operation without human intervention, reducing maintenance needs and improving reliability. It has the advantages over sparking, wear and tear, or toppling these small MCs over contact-based charging methods. Thus, this study proposes a wireless charging system for MCs.

Despite the substantial advancements in the development of MCs and related technologies, several critical research gaps remain unaddressed: (i) Past developments have primarily focused on integrating IoT, machine learning, wireless charging technologies, and path planning as separate components to enhance efficiency and detection capabilities. However, challenges remain in creating comprehensive solutions that combine fast and efficient charging with seamless path optimization to minimize travel time, distance, and energy consumption under dynamic environmental conditions. (ii) Although various algorithms and methods have been proposed for path planning and wireless charging, there is limited research on the real-time adaptability of MC charging in complex and unstructured environments. (iii) The mobile robot platform should have a compact design and a minimal turning radius to enhance maneuverability in urban terrain navigation. (iv) Lastly, existing solutions often do not address the integration of predictive algorithms with continuous learning for more robust and adaptive MC operations.

The key research objectives of the current work are divided into three parts:

• Design and development of a mobile robot platform “Boa” equipped with a fumigation sprayer, utilizing a differential drive system to achieve high maneuverability and flexibility. Also, construction of an MC that uses ultraviolet (UV) light and pheromones to attract mosquitoes, with a suction fan mechanism to trap them securely onto a glue mat.
• Incorporation of an IoT-based, wireless charging MC that can be autonomously charged using the proposed Boa fumigator.
• Development of a GA with an A*-based prioritized path planning algorithm to enable efficient Boa fumigator navigation for charging the MC.

The key novelties of this work involve the following:

➢

The fusion of three distinct techniques—(i) IoT-based charge monitoring, (ii) wireless charging, and (iii) optimized path planning techniques of A*—to enhance the efficiency and real-time application of mosquito control operations without human intervention.

➢

Additionally, this work incorporates a GA with A* path planning algorithm to prioritize MC charging sequence based on the charge percentage and distance. This approach maintains the MC above the minimum charge threshold in autonomous mosquito control and sets this study apart from existing solutions.

This paper follows a structured format to comprehensively address the development and evaluation of fumigation robots and MC in landscaping applications. Section 2 investigates the development of the fumigation robot mobile platform with autonomous capabilities and MC. Section 3 focuses on the development of a prioritized path planning algorithm. Section 4 encompasses a detailed discussion of the path planning algorithm’s results and wireless charging. Finally, in Section 5, this paper summarizes key findings, highlighting the significance of the fumigation robot.

2. Development of Boa Fumigator and Mosquito Catcher

In this study, a mobile robot platform named Boa is developed as the primary unit for autonomous mosquito control operations. Boa is equipped with an integrated fumigation system and serves as a multifunctional robotic platform, as presented in part (1) of Figure 1. The Boa fumigator serves a dual purpose: (i) wirelessly charging a fleet of five MC, as illustrated in part (2) of Figure 1, and (ii) performing fumigation at three designated locations. Parts (3) and (4) of Figure 1 show the MC design, which attracts and traps mosquitoes using UV light and pheromones onto a glue mat. The Boa fumigator autonomously navigates to eight waypoints, starting from its home position W0, and uses GA with A* based prioritized path planning to charge the MC and fumigate. The development of the complete unit is divided into two parts: a Boa fumigator robot and a wireless charging MC device.

2.1. Development of Boa Fumigator Robot

The Boa robot’s locomotion is based on a differential drive, facilitating easier maneuverability on semi-automated planar surfaces in landscaping infrastructure. The robot uses 3D Velodyne, a 2D sick LiDAR sensor, and an inertial measurement unit to facilitate autonomous navigation in the outdoor environment. The Boa robot is equipped with a fumigation payload. The fumigation payload consists of a spray gun and a chemical tank. The gun’s top and bottom ends are attached to one end of the linear actuator and the top end of the 3D-printed mount. The other end of the linear actuator is attached to the bottom end of the metal shaft. This whole system is coupled onto a DC motor to facilitate pan motion for the gun. The linear actuator will facilitate the tilt motion of the fumigation spray gun. Together, the adjustable gun can rotate 360° and fumigate up to 4.5 m from the ground and 2.5 to 6 m away from the robot. Also, a wireless transmitter coil is set up in front of the robot to charge the receiver coil of the MC wirelessly.

Figure 2 depicts the software architecture of the complete fumigation and charging of the MC using the Boa fumigator robot. An onboard ESP8266 microcontroller controls each MC operation. All the MCs send their vital information, like charge percentage, to the Firebase real-time cloud database using a lightweight message queuing telemetry transport (MQTT) protocol. The database keeps track of all the information on the MC and fumigation locations. The Boa fumigator robot subscribes to this information to generate a prioritized sequence order to navigate autonomously from one point to another to charge the MC and fumigation in the least possible time. All the operations on the robot side to perform navigation, fumigation, and charging of the MC are controlled using a robot operating system (ROS). As the robot is deployed in semi-outdoor and outdoor setups, simultaneous localization and mapping (SLAM) algorithms like high-definition lidar (HDL) graph slam are used to map the environment using PointCloud2 data (Shown in Figure 3b) from Velodyne 16 planes LiDAR and later localized using HDL localization for precise accuracy. The 3D map constructed is projected into a 2D plane using the octomap server to obtain a 2D map Figure 3d to facilitate the navigation using the ROS move base nav-stack.

2.2. Development of Wireless Charging Mosquito Catcher

The proposed MC has a Wi-Fi-based microcontroller with an in-built rechargeable LiPo battery, a 5V battery charger, and a wireless charger receiver-rectification circuit. The internal component specifications of the modules are presented in

Table 1. Specifications of Boa fumigator robot, mosquito catcher, and wireless charging unit.

Parameter	Specifications of Boa	Parameter	Specifications of Wireless Charging Circuit
Fumigation Unit	5 L, 360° Rotation, 4.5 m Range		Tx Coil	Rx Coil
CPU Processor, RAM	Intel® Core™ i7-10700 CPU @ 2.90 GHz × 16, 64 GB	Number of turns (N)	12	6
OS	Ubuntu 20.04.6 LTS	Inner Diameter (Din)	60 mm	73 mm
3D LiDAR	Velodyne PUCK 16 plane LiDAR	Outer diameter (Dout)	85 mm	80 mm
2D LiDAR	SICK TiM—571	Width of conductor (w)	1 mm	1 mm
Battery	48 V, 25 Ah	Pitch (p)	0.1 mm	0.1 mm
Brushless Oriental Motor	BLHM450K-GFS	Coil Self-inductance (L)	16.9 µH	5.2 µH
Brushless Motor Driver	BLH2D50-KR	Operational Frequency (f)	141 kHz	141 kHz
IMU	Vectornav VN-100	Power	30 V, 2 A (Max.)	5 V, 6A (Max.)
Parameter	Specifications of Mosquito Catcher	Battery Lifetime Calculations
Microcontroller	ESP8266, 3.3 V, 170 mA (Max.), 15 µA during deep sleep	Usage Duration	Actual Hours
Li-Ion battery	4.2 V, 3400 mAh, 2C discharge	15 min every hour	12 h and 30 min
Li-Ion Charger	5 V, 2 A charge	30 min every hour	8 h and 23 min
Ultraviolet light	5 V, 450 mA	45 min every hour	5 h and 44 min
DC fan	5 V, 130 mA	Continuous UV light, Fan ON every alternative 30 min	4 h and 41 min
DC-DC Buck Converter	60 W, 15 V, 4 A	Continuous Operation	4 h and 21 min

. Studies on insect attraction have shown that the majority of insects are attracted to UV light [29]. Thus, the proposed MC is fitted with a band of UV LEDs. This MC uses a suction fan to trap the mosquitoes onto a glue trap, effectively sticking them onto the surface. To further increase the number of mosquitoes trapped by the catcher, mosquito attraction oil or pheromones are poured onto the top of the MC, as this has been shown to attract mosquitoes [30].

The battery’s working duration determines the MC that must be charged in the proposed path planning algorithm. For the proposed system, let I_T be the total current consumed by the components during its active (from T_ON1 to T_OFF1) and sleep (from T_OFF1 to T_ON2) state, given by Equation (1) and shown in Figure 4. Let I_A be the current consumed by the components during its active state, I_s be the current consumed during the deep-sleep state, I_Fan be the current consumed by the suction fan, I_UV be the current consumed by the UV light, I_MC be the current consumed by the microcontroller, and I_Loss be the losses due to heat and data transmission. During the deep-sleep operation, the suction fan and the UV light will be off-state. Thus, the I_s will be equal to I_MC, given by the following equations:

$$I_T={\int }_{T_{ON1}}^{T_{OFF1}}{I}_A dt+{\int }_{T_{OFF1}}^{T_{ON2}}{I}_S dt+I_{Loss}$$

(1)

$$I_A=I_{MC}+I_{UV}+I_{Fan}$$

(2)

$$I_S=I_{MC}$$

(3)

For the proposed 3400 mAh battery, I_A is 750 mA, I_S is 15 µA, and I_Loss is taken as 120 mA. Considering that the system runs for 15 min every hour and sleeps 45 min, the average consumption is 217.51 mA; thus, the battery life is approximately 12 h and 30 min. The battery is operated for 80% of its full charge capacity, i.e., 80% of 3400 mAh, which is 2720 mAh (20% safety factor). Thus, the percentage decrease per hour is given by 100%/12.5h = approximately 8% decrease in charge per hour.

ANSYS Maxwell simulation case studies were performed to analyze the 3D magnetic field between the transmitter coil (located in the Boa fumigator robot) and the receiver coil (MC), as illustrated in Figure 5. The technical specifications of the coils used for simulation and development are listed in Table 1.

Commercially available wireless charging modules were initially tested to address the vital need to charge MCs efficiently within the proposed autonomous system. However, they failed to meet the operational demands of robotic applications, requiring modifications such as upgraded MOSFETs and custom protection circuits for enhanced thermal management and surge protection. These enhancements ensured reliable long-term performance under dynamic charging conditions. The next chapter elaborates on the formulations, assumptions, algorithms, and implementation of a prioritized path planning algorithm. The algorithm is crafted to prioritize the optimization of the distance between the Boa fumigator robot and the prevailing charge state of the MC. This strategic approach ensures effective utilization of resources and contributes to the system’s seamless functionality and sustained autonomy. Exploration of this path planning algorithm results in a sophisticated solution that maximizes the efficiency of the charging process.

3. Prioritized Path Planning Algorithm

3.1. Problem Formulation

In this section, the problem scenario is mathematically presented to find the optimal path to fumigate an area in different zones and wirelessly charge MC in the same area. The following parameters are set:

• An area environment E: each with waypoints for fumigation and wireless charging.
• Waypoint set W: W  =  {W₁, W₂, …. W_k, W_k+1, ….W_k+l}, where

  ○

  W_k: Charging waypoints;

  ○

  W_l: Fumigation waypoint.

• Activity set A =  {A₁, A₂, …. A_k, A_k+1, …. A_k+l}, which represents a set of activity types associated with each waypoint to indicate whether the robot needs to do fumigation or charging in the corresponding waypoints in W.
• Charge/fumigation value set CF: Represents the initial charge percentage or fumigation time requirements for each waypoint, where CF = {cf₁, cf₂,….cf_k, cf_k+1, …. cf_k+l}.
• Mission sequence seq: Generated sequence based on the current charge levels of the MCs.

Objective: Determine an optimal path (P) and total time (T_o) to complete the fumigation and charging tasks, minimizing the time required.

3.2. Assumptions

For the proposed system, the following list of assumptions are considered:

•

Ÿ Collision-free paths are assumed between waypoints.

•

Ÿ The robot travels at a constant speed from one node to another.

•

Ÿ The robot has sufficient energy to charge all MC and return.

•

Ÿ Linear charging rate for charging individual nodes.

•

Ÿ The discharge rate is significantly less than the charging rate.

3.3. Evaluation of the Mission Sequence

The cost function (X) is defined to minimize the travel and operation times across all waypoints in the mission sequence. It is given by Equations (4) and (5):

$$X ={\mathsf{\Sigma }}_{n=1}^N\left(TravelTim{e}_{\left(n-1, n\right)}+H_n\right)$$

(4)

where

○

TravelTime_(n−1,n): Time to travel from waypoint W_n−1 to waypoint W_n.

○

H_n: Time spent fumigating or charging at W_n.

$$H_n = \left\{\begin{array}{cc}H\left(D_{cf}\left({\mathsf{\Sigma }}_{m=1}^{n}TravelTim{e}_{\left(m-1,m\right)} + H_{m-1} , CF\left[n\right]\right) , C_{max}\right), \hfill & for n>1\hfill \\ H\left(D_{cf}\left(TravelTim{e}_{\left(\mathrm{0,1}\right)}, CF\left[1\right]\right) , C_{max}\right), \hfill & for n=1\hfill \end{array}\right.$$

(5)

○

H(chrg, C_max) is a function that estimates the charge time to reach C_max from chrg or fumigation time in seconds for the considered node.

○

D_cf(chrg, t) is the function to estimate the discharged MC charge percentage from chrg for a time t.

The GA efficiently explores numerous waypoint sequences to find the mission path that minimizes total time. It calculates travel time, fumigation, or charging duration, and predicts each MC discharge as the robot progresses, balancing travel, charging, and fumigation. This approach ensures the selection of an optimal, energy-efficient mission sequence. The pseudocode in Algorithm 1 depicts the evaluation of the mission sequence considered. Consider an environment where W, A, and CF are described in the problem formulation. The mission sequence (seq) generated by the optimization considers only the waypoints with MC charge below a fixed threshold level for that zone. Different combinations of mission sequences of waypoints are generated, and the algorithm evaluates them to find the path and total time.

The algorithm iterates over the waypoints in seq sequentially. Each iteration estimates an A* collision-free path between the robot’s current position and the goal point w_g, considered from seq. The path length is calculated for the estimated path to find the robot’s travel time to reach w_g. Furthermore, the algorithm predicts the discharge percentage for all other MC while traveling to w_g by using D_cf. After reaching w_g, depending on the activity type of w_g, the robot charges/fumigates at the node for a certain amount of time dth, which is calculated using H. During this process, there is discharge on other MCs for dth time. Thus, D_cf is used again to estimate the discharges in other waypoints in seq. The total time T_o is updated with T_o + cycle_time, where cycle_time is dt  +  dth. Next, the robot’s current position is updated to w_g, w_g to the next waypoint in seq, and CP is added with the new path generated in this iteration. This process is iterated until all the waypoints in the seq are completed and restarted with new nodes to charge.

Algorithm 1 evaluates the mission sequence to find the optimal path and time to complete the mission. So, an optimal mission sequence is required to find the optimal solution. A repair function is used with the GA to generate different combinations of waypoints considered from W. This repair function ensures that seq always starts from the robot position. Figure 6 shows the GA flowchart, where the repair function is used to generate sequences, and Algorithm 1 shows the process flow in the evaluation stage of the optimization.

Algorithm 1: Estimation of optimal path and total time taken to complete fumigation and wireless charging of MC
Inputs:
W A N m Rp seq CF	: Waypoint coordinates  $W=\left\{W_1,W_2 ,\dots ., W_k ,W_{k+1 }, \dots .W_{k+l} \right\}$ : Activity at the waypoint (Fumigation or Charging) $A=\left\{A_1,A_2 ,\dots ., A_k , A_{k+1 }, \dots .A_{k+l} \right\}$ : Number of charging waypoints : Number of fumigation waypoints : Current Robot Position : Sequency of charging : MC charge values or Fumigation values CF $=\left\{c{f}_1,c{f}_2 ,\dots ., c{f}_k , c{f}_{k+1 }, \dots .c{f}_{k+l} \right\}$
Output:
Po To	: Output path : Total time taken to complete the path
Initializations:
CFCurrent Rp-current wg CP Plen ds dt Cmax To	← CF, where CFCurrent is the current MC charge values & fumigation values ← Rp, Rp-current is the current robot position. ← goal waypoint taken from seq ← 0, Complete path of the robot ← 0, ← 0, (robot travel time) ← 0, ← 90%; ← 0
Functions:
Lp(P) Dcf(cfvalues, to), to H(cfvalues, Chargmax)		: Length of the path P : Function to calculate discharge of MC after to : Function to estimate time to fumigate/charge MC
Pseudocode:
Extract seq from optimization after considering the charge levels of each MC. wg is the first waypoint from the seq
while (wg exists) do:
Path plen dt cfn1 dth cfn2 CFcurrent To Rp-current CP wg	← find path from Rp-current to wg ←Lp(Path) ← ds/Rvel ←Dcf(CFCurrent, dt) ←H(cfn1(wg), Cmax) ←Dcf(cfn1, dth) ← cfn2 ← To + dt + dth ← wg ← Path ← Select next waypoint from seq array
END while

4. Results and Discussion

The Section 4 is divided into two sub-sections: results pertaining to the prioritized path planning algorithm for MC charging test cases and results of power transfer achieved during the wireless charging of the MC by the fumigation robot.

4.1. Prioritized Path Planning

The prioritized path planning sequence between the nodes to be charged is explained via three test cases. In these three cases, five charging nodes and three fumigation nodes are considered (a total of eight waypoints). The robot starts from its home position (W0) to perform charging and fumigation. The activity values 1 and 2 represent fumigation and charging, respectively. In all three cases, W4 is a charging node whose initial charge percentage of the MC is above 90%. The optimization only considers the fumigation and charging nodes with an initial charge percentage of less than 90%.

A.

Case 1: When the nearest MC has high priority

This case depicts a scenario where the path is prioritized to charge a node with the least charge and which is near the robot-initialized position. From

Table 2. When the nearest MC has high priority.

Waypoint (W)	A* Distance from W0 (m)	x	y	Activity	Charge%
0	0	0.16159	−1.30728	-	-
1	4.15	16.04708	5.66740	1	F
2	7.19	24.81930	16.20937	2	75
3	10.38	13.01278	20.02445	1	F
4	9.21	−0.41344	28.67112	2	93
5	7.61	−1.85558	20.76541	2	55
6	4.79	−6.29833	10.31836	1	F
7	4.45	−12.31793	5.25405	2	25
8	7.19	−15.64924	19.36002	2	45
		Prioritized Path		0, 7, 6, 8, 5, 3, 1, 2

Note: “1” in activity represents fumigation activity (F); thus, it is not considered for wireless charging. The number “2” in activity represents charging the mosquito device, and their respective charge percentages are shown in the Charge% column. Furthermore, charging or fumigation does not apply to W0 (Boa’s home position).

, W7 has the least charge, followed by W8, W5, and W2. W4 is not considered, as the initial charge exceeds 90%. As W7 has the least charge and is near the robot’s initial position, the optimization prioritizes moving the robot towards it to charge the MC. Later, it predicts that the least charge is in W8, but it moves to W6 to fumigate rather than W8 to charge because traveling time + charge time + fumigation time is less in W7→W6→W8→W5 than in W7→W8→W6→W5, which accounts for the total time to complete the mission.

B.

Case 2: When the farthest MC has high priority

This case depicts a scenario where the path is prioritized to charge a node with the least charge, far away compared to other waypoints from the robot’s initial position.

Table 3. When the farthest MC has high priority.

Waypoint (W)	x	y	Activity	Charge%
0	0.16159	−1.30728	-	-
1	16.04708	5.66740	1	F
2	24.81930	16.20937	2	25
3	13.01278	20.02445	1	F
4	−0.41344	28.67112	2	93
5	−1.8555	20.76541	2	60
6	−6.29833	10.31836	1	F
7	−12.31793	5.25405	2	65
8	−15.64924	19.36002	2	45
Prioritized Path		0, 2, 3, 5, 8, 6, 7, 1

shows that W2 has a very low charge, followed by W8, W5, and W7. The optimization has prioritized W2 to charge to avoid going into a dead state. Also, the greater the discharge of any MC, the greater the time it takes to charge the MC, which leads to a greater total time needed to complete the mission. So, it prioritizes W2 and charges and predicts charges for other MC. After the prediction of charges of other waypoints, instead of considering W8 to charge first, it considers W2→W3→W5→W8 because the total time cost is greater in charging W8 first and later fumigating and charging W5 than in the former case. Also, W2→W3→W5→W8 does not allow for W8 to go into a dead state, which cumulatively leads to the path consuming less time. Later, it follows W8→W6→W7→W1 to complete the mission.

C.

Case 3: When multiple MCs have high priority.

This case depicts a scenario where two nodes have high priority because of low and almost equal charges. Once again, W4 is not considered because the initial charge percentage is above the threshold.

Table 4. When multiple MCs have high priority.

Waypoint (W)	x	y	Activity	Charge%
0	0.16159	−1.30728	-	-
1	16.04708	5.66740	1	F
2	24.81930	16.20937	2	70
3	13.01278	20.02445	1	F
4	−0.41344	28.67112	2	93
5	−1.8555	20.76541	2	36
6	−6.29833	10.31836	1	F
7	−12.31793	5.25405	2	65
8	−15.64924	19.36002	2	35
Prioritized Path		0, 8, 5, 6, 7, 2, 3, 1

shows that W8 and W5 have the lowest charge percentages, followed by W7, W2, and W4. As W8 is closer to the robot’s initial position than W5, the algorithm prioritizes charging W8 first, then W5. The next one in the priority order comes the W2, but the robot chooses to fumigate W6 and then move towards W7 to account for less travel time. Later, the robot moves in the order W7→W2→W3→W1 to complete the mission. Figure 7 shows the path generated by the algorithm for all three cases.

4.2. Robot-to-Mosquito Catcher Power Transfer

From the simulation results, as illustrated in Figure 8a, it is observed that, at the 0 mm lateral displacement between the transmitter coil and the receiver coil—i.e., both the coils are placed concentric to each other—a stronger magnetic field is generated. This alignment maximizes the flux linkage from the transmitter to the receiver coil, resulting in maximum energy/power transfer from the Boa fumigator robot to the MC. In contrast, a weaker magnetic field is produced when the transmitter and the receiver coils are placed with a lateral displacement of 20 mm, as shown in Figure 8b. This displacement reduces the flux linkage between the transmitter and receiver, significantly decreasing energy transfer efficiency to the MC.

The results in Figure 9 indicate the wireless charging performance of the fumigation robot and the MC at various distances. When the input voltage is 25–30 V, the output voltage at a 10 mm distance is measured at 22 V with a maximum current of 2 A. As the distance increases to 15 mm, the output voltage decreases to 20 V. The observed voltage and current values suggest that the wireless charging efficiency decreases with increased distance between the fumigator robot and the MC. This phenomenon can be attributed to electromagnetic field attenuation and energy loss over distance. The wireless charging system is most efficient at a proximity of 10 mm, as indicated by the higher output voltage and current. However, as the distance between coils increases, the electromagnetic field linking the receiver coil diminishes. Figure 10 illustrates the output voltage values at different lateral displacements between the transmitter and receiver coils. At a constant 30V input to the transmitter coil, an output of 25 V is received from ±10 to 15 mm; however, the power transfer is still acceptable until 30 mm displacement for low-power devices.

In line with the simulation results and experimental results presented in Figure 9 and Figure 10, maximum energy transmission can be achieved within a range of −15 mm to +15 mm lateral displacement, indicating the importance of maintaining alignment for effective power transfer.

4.3. Proposed System Use Case and Advantages

An autonomous fumigation robot and IoT-powered MC offer a targeted solution for mosquito population control while addressing the challenges in traditional methods. Some notable use cases are as follows:

•

Ÿ Precision targeting of mosquito breeding hotspots through intelligent surveillance and data analysis.

•

Ÿ Optimized deployment of fumigation chemicals, ensuring efficient area coverage in breeding hotspots and quantified use of chemicals based on mosquitoes captured in specific hotspots.

•

Ÿ Wireless charging makes the MC system less prone to damage due to fumigation chemicals and elements such as water and dust.

•

Ÿ Targeted fumigation based on real-time mosquito density minimizes the use of chemicals and reduces the environmental impact associated with widespread spraying.

•

Ÿ The proposed system is scalable for deployment in various environments, including urban areas, agricultural fields, and public spaces.

•

Ÿ Multiple robots can work collaboratively, covering larger areas and addressing diverse mosquito breeding grounds.

•

Ÿ Wireless IoT-based connectivity allows for remote monitoring of the fumigation robot, allowing for operators to be more flexible and adaptable to different scenarios.

4.4. Limitations and Future Scope

The limitations of this study and the corresponding future scopes are outlined as follows:

•

Ÿ The battery life of the Boa robot is not explicitly considered in this study. Future work will explore energy optimization strategies and incorporate battery life analyses into the system design.

•

Ÿ All the experiments were conducted exclusively in urban terrain settings. Conditions such as irregular surfaces and slippery and inclined terrains were not considered, which can cause the Boa robot to slip. Future research will test the robot’s performance in diverse terrain conditions and improve its adaptability to such challenges.

•

Ÿ This study focuses on a single Boa robot with five MCs. Scaling it to larger fleets and more fumigation points may face coordination, power, and path optimization challenges, leading to the deployment of multiple fumigation robots. Future work will develop advanced coordination algorithms to handle multi-robot systems effectively.

•

Ÿ The performance of the Boa robot and MC was tested under ideal conditions, which may be affected by extreme weather conditions (e.g., rain, extreme heat). Future studies will examine the robot’s robustness and efficiency under varying environmental conditions.

•

Ÿ While commercially available wireless charging modules were tested, they were inadequate for long-term sequential charging, resulting in overheating and component failures. A customized wireless charging system was developed for this study, but its performance was not analyzed in detail. Future work will focus on designing and refining the custom charging system and conducting a detailed performance analysis to further its efficiency and reliability.

5. Conclusions

This paper presents an autonomous fumigation and charging system, integrating a prioritized path planning algorithm with a wireless charging system for efficient mosquito population control. The proposed path planning algorithm leverages GA to optimize the robot’s mission sequence, prioritizing waypoints based on charge levels and proximity, ensuring that the MC devices maintain a charge above 30% throughout the operation. With this prioritized approach, the MC devices achieve a battery life of up to 12 h and 30 min. The optimized path ensures the timely recharging of each MC device while allowing for the robot to complete fumigation tasks effectively. Experimental results show that the path planning algorithm adapts well to different displacement tolerances, supporting consistent performance within the charging range. A wireless charging system was developed to transmit 22 V and 2 A to the MC receiver. It demonstrates robust performance within a displacement range of −15 mm to +15 mm, with a maximum functional displacement of 30 mm for the current application. This stability in wireless power transfer allows for the robot to maintain optimal charging performance even with slight positional variations. Future work will refine the path planning algorithm to accommodate more dynamic environmental conditions and enhance wireless charging efficiency for increased operational resilience.