Dual-Unloading Mode Autonomous Operation Strategy and Cotransporter System for Rice Harvester and Transporter

Fan Ding; Xiwen Luo; Zhigang Zhang; Lian Hu; Xinluo Wu; Kaiyuan Bao; Jiarui Zhang; Bingxuan Yuan; Wenyu Zhang

doi:10.1016/j.eng.2024.11.006

Engineering ›› 2025, Vol. 48 ›› Issue (5) :233 -246. DOI: 10.1016/j.eng.2024.11.006

Research Agricultural Machinery—Article

research-article

Dual-Unloading Mode Autonomous Operation Strategy and Cotransporter System for Rice Harvester and Transporter

Fan Ding ^a^,^c
, Xiwen Luo ^a^,^b^,^c
, Zhigang Zhang ^a^,^b^,^c
, Lian Hu ^a^,^b^,^c
, Xinluo Wu ^a^,^c
, Kaiyuan Bao ^a^,^c
, Jiarui Zhang ^a^,^c
, Bingxuan Yuan ^a^,^c
, Wenyu Zhang ^a^,^b^,^c^,^*

Author information +

History +

PDF (4639KB)

Abstract

To achieve an unmanned rice farm, in this study, a cotransporter system was developed using a tracked rice harvester and transporter for autonomous harvesting, unloading, and transportation. Additionally, two unloading and transportation modes—harvester waiting for unloading (HWU) and transporter following for unloading (TFU)—were proposed, and a harvesting–unloading–transportation (HUT) strategy was defined. By breaking down the main stages of the collaborative operation, designing module-state machines (MSMs), and constructing state-transition chains, a HUT collaborative operation logic framework suitable for the embedded navigation controller was designed using the concept and method of the finite-state machine (FSM). This method addresses the multiple-stage, nonsequential, and complex processes in HUT collaborative operations. Simulations and field-harvesting experiments were performed to evaluate the applicability of this proposed strategy and system. The experimental results showed that the HUT collaborative operation strategy effectively integrated path planning, path-tracking control, inter-vehicle communication, collaborative operation control, and implementation control. The cotransporter system completed the entire process of harvesting, unloading, and transportation. The field-harvesting experiment revealed that a harvest efficiency of 0.42 hm²∙h⁻¹ was achieved. This study can provide insight into collaborative harvesting and solutions for the harvesting process of unmanned farms.

Graphical abstract

Keywords

Agricultural machinery / Harvesting–unloading–transportation strategy / Cotransporter system / Unmanned farm / Finite-state machine

Cite this article

Download citation ▾

Fan Ding, Xiwen Luo, Zhigang Zhang, Lian Hu, Xinluo Wu, Kaiyuan Bao, Jiarui Zhang, Bingxuan Yuan, Wenyu Zhang. Dual-Unloading Mode Autonomous Operation Strategy and Cotransporter System for Rice Harvester and Transporter. Engineering, 2025, 48(5): 233-246 DOI:10.1016/j.eng.2024.11.006

登录浏览全文

4963

注册一个新账户忘记密码

1. Introduction

The rapid development of agricultural mechanization, informatization, and intelligent technology has led to the emergence of unmanned farm technology, attracting considerable attention from researchers [1], [2]. Unmanned farms rely on advancements in autonomous driving technologies for agricultural machinery. They operate in an all-weather, full-process, and fully automated mode. Unmanned farms are capable of completing every aspect of farm production and management tasks autonomously. They are the most advanced agricultural productivity approach that can significantly improve labor productivity and resource utilization.

During the harvesting process of an unmanned farm, multiple rounds of unloading and transportation are required due to the limited capacity of the grain tank of the harvester, thus reducing the efficiency of harvesting. Therefore, a collaborative operation strategy between the harvester and transporter must be developed to fully achieve autonomous harvesting, unloading, and transportation to improve the operational efficiency of a harvester. This strategy can serve as theoretical support and technical reference for multivehicle collaboration (e.g., reseeding and refueling) and other operation scenarios.

1.1. Key technologies of harvesting–unloading–transportation (HUT) collaborative operation

The complexity of a collaborative harvesting operation is relatively high due to its multiple aspects, such as path planning and operation scheduling, path-tracking control, inter-vehicle communication, collaborative operation control, and implementation control. Path-tracking control is the basic technology of intelligent agricultural machinery equipment and has been extensively investigated. The control methods include the pure pursuit algorithm [3], [4], proportional–integral–derivative (PID) control [5], [6], fuzzy control [7], [8], sliding-mode control [9], [10], and model predictive control [11], [12]. The accuracy of tracking control can reach ± 2.5 cm. Inter-vehicle communication serves as the foundation of information exchange between the harvester and transporter. It can be achieved through technologies including radio [13], the 4th generation mobile communication technology (4G) [14], bluetooth [15], and long-range radio (LoRa) [16]. Inter-vehicle communication can also be achieved by combining communication protocols and data-processing methods.

Path planning provides a benchmark for making decisions. Several researchers have begun to investigate path planning and operation scheduling related to multimachine collaborative harvesting operations. Yao et al. [17] conducted a path-planning study on the collaboration of the same type of agricultural machinery. They simulated multimachine conflict-free collaborative operation with path optimization to optimize the total operation time. The simulation results showed that the total operation time of a rectangular farmland decreased by an average of 2.45% and 2.29%. Zhai et al. [18] developed a safety-state detection model for agricultural machinery based on the directional bounding box algorithm and separation axis theorem. They proposed a multimachine collaborative navigation path-planning method for the master–slave following mode. This method can provide global operation paths for both master and slave machines, thus effectively preventing collisions between them during turns. Zhang et al. [19] proposed a multimachine collaborative operation task-scheduling method using an improved genetic algorithm to simultaneously allocate multiple agricultural machinery stations and meet the real-time operation orders of farmers. Wang et al. [20] adopted an improved Dijkstra algorithm using priority queues to achieve full-process path planning for a single agricultural machine and multiple agricultural machines of the same type. They verified the proposed algorithm through simulation.

In recent years, researchers have conducted relevant studies pertaining to collaborative operational control. Noguchi et al. [21] first proposed the “GOTO” and “FOLLOW” algorithms using the master–slave agricultural multi-robot system. These proposed algorithms are widely applicable to agricultural operation scenarios. Using the “FOLLOW” method, proportional–derivative (PD) control, and sliding-mode control, they designed the following collaborative operation control method for master–slave agricultural robots. Their simulation test results revealed that the maximum root mean square error of the relative lateral and longitudinal positions between the master and slave was 0.134 m. Zhang et al. [13] adopted a PD control method combining state feedback and disturbance feedforward to enable a semiautomatic agricultural vehicle (slave) to follow a leading tractor (master) at a specified lateral and longitudinal offset. Zhang et al. [15] used a PD controller to maintain the distance between two vehicles and proposed a safe, collaborative turning method. Field test results showed that at a tractor speed of 0.83 m∙s⁻¹, the average error of the relative position of two vehicles was 0.13 m, and the variance was 0.15 m. Bai et al. [22] proposed a corporative navigation control strategy for a harvester group using the leader–follower structure. They combined feedback linearization, sliding-mode control theory, and a harvester group kinematics model to design a path-tracking and formation-preserving control method. The test results showed that the average errors of the leader and follower were 0.058 and 0.059 m, respectively. Li et al. [23] designed a fuzzy controller based on lateral and heading deviations, which achieved the slave-following-master pattern using the path provided by the master. The root mean square error of the path tracking at a slave speed of 0.8 m∙s⁻¹ was 0.068 m. Zhang et al. [24] proposed a longitudinal relative-position control method with position–velocity coupling to realize a collaborative harvesting–unloading operation for wheeled harvesters and transport tractors. At a harvester speed of 1 m∙s⁻¹, the maximum standard deviation of the longitudinal relative-position deviation was 0.091 m. Ding et al. [25] proposed a self-adjusting single-neuron PID control method to accurately control the longitudinal relative position of a tracked harvester and tracked transport vehicle. Field-harvesting experiments showed that the steady-state maximum deviation was 0.253 m and that the grains were accurately unloaded into the granary of the transport vehicle.

In summary, significant progress has been made in studies on path-tracking control, inter-vehicle communication, and collaborative operation control. Although these methods have some disadvantages, they can satisfy the requirements of collaborative harvesting operations. Meanwhile, path planning and operational scheduling have been investigated through simulation. Although the results provide an excellent theoretical foundation and valuable insights for solving practical problems, they cannot be directly applied to practical applications. Therefore, to achieve fully autonomous harvesting and enhance operational efficiency, it is crucial to integrate path planning and operation scheduling, path-tracking control, inter-vehicle communication, collaborative operation control, and implementing control based on actual production demands and operational scenarios. This approach will help develop the HUT collaborative operation strategy suitable for actual production and operation.

1.2. Research objectives

This study aimed to design a HUT method and strategy that combined path planning, path-tracking control, inter-vehicle communication, collaborative operation control, and implementing control in the spatial and temporal sequences of the collaborative harvesting operation. This combined strategy aimed to achieve the entire process of autonomous HUT operation and provide support—from theoretical models to technical implementation—for master–slave collaborative harvesting.

In this study, a full process autonomous HUT mode that combines two different grain-unloading methods, harvester waiting for unloading (HWU) and transporter following for unloading (TFU), was first proposed. Then, based on the concept and method of the finite-state machine (FSM), a HUT collaborative operation logic framework suitable for embedded navigation controllers was designed by breaking the main stages of the collaborative operation, designing module-state machines (MSMs), and constructing state-transition chains. Finally, simulations and field harvest experiments were conducted to validate the HUT collaborative operation strategy.

2. Materials and methods

2.1. Experimental equipment and materials

2.1.1. Experimental vehicles

In this study, an RG60V4G-036 tracked harvester and an RG60V4G-037 tracked transporter (WEICHAI LOVOL, China) were used as the experimental vehicles. Both the harvester and transporter were equipped with electrically controlled chassis, enabling electric control of hydraulic steering, hydraulic stepless transmission, engine, header, threshing drum, grain-unloading cylinder, and other actuators. The main technical parameters of the harvester and transporter are listed in Table 1.

2.1.2. Overall design of the cotransporter system

The designed cotransporter system mainly comprised a harvester, a transporter, positioning and attitude measurement modules, inter-vehicle communication modules, and navigation and autonomous operation control modules. Both vehicles used a dual-antenna global navigation satellite system (SinoGNSS K726; ComNav Technology Ltd., China) at a positioning information acquisition frequency of 10 Hz. The horizontal positioning accuracy of the system was ± (10 + 1 × 10⁻⁶ × D) mm, where D is the distance from the base station to the mobile station (km). Inter-vehicle communication was developed using a 4G-cat1 data transmission terminal (USR-DR152; Jinan USR IOT Technology Ltd., China) and a 2.4 GHz full-duplex wireless data transmission radio modem (AS69-DTU20; Chengdu Ashining Technology Ltd., China), and their corresponding data transmission frequencies were 2 and 5 Hz, respectively. STM32F407ZGT6 minimum system boards (ALIENTEK-STM32F407ZGT6 minimum system boards; Guangzhou Xingyi Electronic Technology Co., Ltd., China) were used for data processing and the fusion module. The navigation and autonomous operation control terminal comprised a STM32H743IGT6 as the core processor, which communicated with the chassis electronic control unit of the vehicles through a controller area network (CAN) bus. The software for the control terminal was developed and debugged using Metrowerks Code Warrior for ARM Developer Suite v1.2. The hardware composition and structural diagram of the tracked cotransporter system are shown in Fig. 1.

2.2. Definition of HUT collaborative operation

2.2.1. Dual-unloading mode of collaborative unloading

Noguchi et al. [21] from Hokkaido University first proposed the “GOTO” and “FOLLOW” algorithms. These algorithms are widely applicable to agricultural machinery operations. Based on the actual production demands and operation environment of unmanned farms, in this study, two types of collaborative unloading methods were proposed: HWU and TFU.

In the HWU system, the harvester stops and summons the transporter when its granary of the harvester reaches a designated threshold, and unloading occurs after the transporter is situated in its intended location. In the TFU system, the transporter is summoned to follow in the straight section of the harvest operation when the granary of the harvester reaches a designated threshold, and two vehicles collaborate to accurately complete the unloading operation. Both HWU and TFU can achieve complete autonomous unloading and transportation. The operational scenario is illustrated in Fig. 2.

TFU allows the harvester to unload without stopping, thereby significantly improving the efficiency of the operation. However, it is not applicable to all scenarios. When the first unloading is required after harvest begins, if the transporter lacks sufficient space to follow the harvester, then HWU must be performed first. Once sufficient space has been cleared in the field, then TFU can be conducted. After the entire field has been harvested, the harvester must unload all grains into the transporter, making HWU more suitable. Therefore, the combination of HWU and TFU was used in this study to achieve autonomous HUT in the entire process.

2.2.2. Primary stages of collaborative operation

The collaborative harvesting mode combining HWU and TFU can be divided into eight primary stages based on the characteristics of collaborative operation and control:

In stage 1 (S₁), the harvester automatically navigates and harvests, tracking the planned path and activating the relevant implementation for autonomous harvesting.

In stage 2 (S₂), the transporter waits for summoning. It drives to the starting point of the collaborative path section of HWU based on the planned path, stops, and waits for the harvester to be summoned to HWU or switch to the TFU mode.

In stage 3 (S₃), the transporter waits for TFU. When it receives the summoning instruction from the harvester, it navigates based on the planned path to the starting point of the collaborative path section of TFU, stops, and waits for TFU.

In stage 4 (S₄), the harvester and transporter execute HWU. When the granary of the harvester reaches a designated threshold and the unloading position, the harvester stops and sends the summoning instruction to the transporter. After receiving the summons, the transporter vertically aligns with the harvester along the planned collaborative HWU path. After the transporter is positioned in its intended location, the harvester begins unloading, achieving HWU.

In stage 5 (S₅), the harvester and transporter execute TFU. When the harvester reaches the collaborative TFU path, it continues to harvest and then sends instructions to the transporter to begin driving. The transporter adjusts its speed in real time and longitudinally aligns with the harvester. After the position is aligned, the harvester begins to unload, achieving TFU. Based on the sensor information of the harvester’s granary, or if the harvester is about to turn, it stops unloading, and then the transporter stops.

In stage 6 (S₆), after the transporter completes HWU. S_6-1: if the next unloading mode is TFU or the granary reaches the threshold, then stage 8 (S₈) is executed for unloading; S_6-2: if the next unloading mode is HWU and the granary does not reach the threshold, then S₂ is executed to wait for the next HWU.

In stage 7 (S₇), after the transporter completes TFU. S_7-1: if the next unloading mode is HWU or the granary reaches the threshold, then S₈ is executed for unloading; S_7-2: if the next unloading mode is TFU and the granary does not reach the threshold, then the transporter waits for summoning. When the transporter receives the summoning instruction from the harvester, it plans the path and drives to the starting point of the collaborative TFU path. After the transporter is situated in its intended location, S₃ is executed to wait for TFU.

In S₈, the unloading stage of the transporter, the transporter plans the unloading path and drives to the designated location for unloading. After unloading, the transporter replans the path and enters S₂ or plans the return path to the garage and returns to the garage.

S₁, S₂, S₃, S₆, S₇, and S₈ are weak coupling stages. The harvester and transporter are only required to maintain inter-vehicle communication and complete their own independent planning and operational tasks. S₄ and S₅ are strong coupling stages that require real-time communication and collaboration between the harvester and transporter to control the relative longitudinal distance and ensure a collaborative operation and accurate unloading.

2.2.3. Process of collaborative operation

The HUT collaborative operation was designed based on the proposed primary stages. The harvester first departs from the garage and drives to the field based on the planned path, executing S₁ for harvesting. Then, the transporter departs from the garage, executes S_2, and waits for a summons. The harvester summons the transporter based on the unloading method and granary information. After receiving the summons, the transporter executes S₄, S₆, S₈, and S₂ to complete HWU and transportation or executes S₃, S₅, S₇, S₈, and S₂ to complete TFU and transportation. After harvesting is completed, the harvester summons the transporter to execute S₄ and unloads all grains into the transporter. The harvester first returns to the garage and then executes S₈ for unloading and returns to the garage. The two vehicles collaborate to complete the HUT collaborative operation. The primary stages and the process of collaborative operation are illustrated in Fig. 3.

2.3. Modeling of HUT collaborative operation

The HUT collaborative operation features multiple stages and nonsequential and complex processes. Adopting conventional control logic and structures leads to complex logic and convoluted designs, thus complicating maintenance and upgrades. The FSM is the research object of automata theory and represents the mathematical model of finite states and behaviors, such as transitions and actions between states [26]. The FSM features the following three characteristics: ① the total number of states is finite, ② at any moment, the state machine is in only one state, and ③ it transitions from one state to another when a specific condition is triggered [27], [28], [29]. FSM has been applied to achieve complex agricultural scenarios, such as autonomous driving of tractors on farmland roads [30] and collaborative control of the cotton picking robot [31]. The advantages of FSM are as follows: ① It can make the framework of control logic concrete and easy to implement; ② the characteristics of the state actuator enable errors to be accurately located; ③ modular design is easy to maintain and upgrade.

According to the concepts and methods of FSM, the actions of the harvester and transporter during the primary stages of the HUT collaborative operation (in Section 2.2.2) can be described as a finite number of states. State transfers can be realized by describing and defining the transfer conditions for each state. Finally, by constructing the state-transition chain and logical framework, the corresponding actions of the harvester and transporter are realized and changed to complete the entire complex HUT collaborative operation.

2.3.1. MSM and state-information matrix

MSMs were established for the harvester and transporter. An MSM contains a state-information matrix and state-function modules. The state-function modules mainly include a state actuator (A), a state evaluator (E), a state trigger (T), and state verification (V). The MSM for the harvester and transporter is denoted as M_h and M_t, respectively. The relationship between them is expressed as follows:

(1)

\{\begin{matrix} M_{h} = {Q_{h}, A_{h}, E_{h}, T_{h}, V_{h}} \\ M_{t} = {Q_{t}, A_{t}, E_{t}, T_{t}, V_{t}} \end{matrix}

where Q_h and Q_t are the state-information matrices of the harvester and transporter, respectively.

The state-information matrix of the harvester can be expressed as

(2)

Q_{h} = [Q_{h 1}, Q_{h 2}, Q_{h3}, . . ., Q_{h 13}]

where Q_h1 is the path-planning state, Q_h2 is the driving-speed state, Q_h3 is the engine-speed state, Q_h4 is the height state of the cutting table, Q_h5 is the main-clutch state, Q_h6 is the cutting-table clutch state, Q_h7 is the unloading-clutch state, Q_h8 is the position state of the grain-unloading barrel, Q_h9 is the granary-monitoring state, Q_h10 is the state of straight-line tracking or turning, Q_h11 is the evaluation state of real-time heading and lateral deviation, Q_h12 is the evaluation state of unloading routes and methods, and Q_h13 is the real-time distance from the transporter.

The state-information matrix of the transporter can be expressed as

(3)

Q_{t} = [Q_{t1}, Q_{t 2}, Q_{t3}, . . ., Q_{t 13}]

where Q_t1 is the path-planning state, Q_t2 is the driving-speed state, Q_t3 is the engine-speed state, Q_t4 is the unloading-clutch state, Q_t5 is the position state of grain-unloading barrel, Q_t6 is the granary-monitoring state, Q_t7 is the evaluation state of longitudinal alignment in HWU (for calculating the longitudinal deviation), Q_t8 is the evaluation state of longitudinal alignment in TFU (for calculating the longitudinal deviation), Q_t10 is the state of straight-line tracking or turning, Q_t11 is the evaluation state of real-time heading and lateral deviation, Q_t12 is the evaluation state of unloading routes and methods, and Q_t13 is the real-time distance from the harvester.

2.3.2. Definition of state-function modules

2.3.2.1. State actuator

The state actuator is used to describe and represent the states of various actuators of the harvester and transporter in the current state. As presented in Section 2.2, all execution actions of the harvester and transporter can be categorized into several fixed basic execution states. Analyzing and modeling these states enable the division and description of the actions of the harvester and transporter at each stage of the HUT collaborative operation. The definitions of state actuators and the state-information matrices in state actuators are listed in Table 2. The corresponding actions of the harvester and transporter can be achieved and changed by sequentially changing the corresponding state-information matrices.

2.3.2.2. State evaluator

The state evaluator describes the evaluation metrics from the internal sources and uses them to determine whether the state actuators initiate state transfer. The definitions of state evaluators and the state-information matrices in state evaluators are listed in Table 3.

2.3.2.3. State verification

State verification is performed to calculate and validate the results of the current state and to determine whether the state actuators initiate state transfer. The definitions of state verifications and the state-information matrices in state verifications are listed in Table 4.

2.3.2.4. State trigger

State trigger describes the external trigger signal. It combines with the internal-state information to trigger state actuators to perform state transfer. The definitions of state triggers and the state verifications associated with the state triggers through inter-vehicle communication are listed in Table 5.

2.3.3. Construction of state-transition chain and logical framework

Based on the definition of state-function modules, the state actuators represent the execution actions of the harvester and transporter at each stage of the HUT collaborative operation. The entire HUT collaborative operation can be described as the state transfer among the various state actuators. State transfer is determined by three functional modules: the state evaluator, state trigger, and state verification. The state-transition chain for the HUT collaborative operation can be designed based on the collaborative operation of HUT described in Section 2.2 (Fig. 4). In M_h: A_h1→A_h2→A_h3-1→A_h3-2→A_h1 are HWU stages; A_h1→A_h4→A_h5-1→A_h5-2→A_h1 are TFU stages; and A_h1→A_h6→A_h1 are temporary parking stages of the harvester waiting for the transporter. In M_t, the state transition chain of the transporter is designed around five parking states, as shown in Fig. 3(b). Q_t1 stands for calling path plan. A_t2-1→A_t1-2→A_t2-3 are HWU stages, A_t2-1→Q_t1→A_t1-1→A_t2-2 represent the transporter driving to the location waiting for TFU, and A_t2-2→Q_t1→A_t1-3→A_t2-4 are TFU stages. At A_t2-3 and A_t2-4, based on the granary information and the next unloading method of the harvester, A_t2-3→Q_t1→A_t1-1→A_t3-1 or A_t2-4→Q_t1→A_t1-1→A_t3-1 are executed for unloading, or A_t2-3→A_t2-1 or A_t2-4→Q_t1→A_t1-1→A_t2-2 are executed to continue receiving grain. After the transporter unloads grain, A_t3-1→A_t3-2→Q_t1→A_t1-1→A_t2-1 are executed to wait for summoning, or A_t3-1→A_t3-2→Q_t1→A_t1-1 are executed to return to the garage. E and V are internal conditions for evaluating whether A performs a state transition; T are external conditions implemented by inter-vehicle communication. Therefore, the HUT collaborative operation described in Section 2.2.3 of this study can be expressed and modeled by the state-transition chain, and it has good module independence and scalability.

The logical framework design of the HUT collaborative operation is crucial for collaborative harvesting, beginning from theoretical modeling to technical implementation. It integrates and effectively matches key aspects, such as path planning, path tracking, inter-vehicle communication, longitudinal position control, and implementation control, thus enabling the cotransporter system to complete the entire process of autonomous HUT operation, which combines HWU and TFU. Based on the state-transition chain of the HUT collaborative operation and the embedded navigation controller of the cotransporter system, a logical framework was designed in the form of repeated retrieval of the main cycle. The logical flow diagrams of the harvester and transporter are shown in Figs. S1 and S2 in Appendix A, respectively. The transmission of trigger signals between two vehicles was achieved through inter-vehicle communication, thus enabling the two vehicles to collaborate and complete HUT operations together.

2.4. Algorithm evaluation and validation

2.4.1. Simulation method

To verify the state-transition chain and logical framework established in Section 2.3, MATLAB Simulink and Stateflow modules were used for simulation and analysis. The logical framework was simplified by replacing the time relationship with the step size; the established collaborative harvesting logical simulation model is shown in Fig. 5, where “harvester” and “transporter” represent the Stateflow modules for each respective system. Delays 1 and 2 represent the durations for executing the current action. Delays 3–7 represent the inter-vehicle communication delays. The collaborative unloading method information provided by harvester path planning (ULM). Two instances of HWU and three instances of TFU were designed in the simulation. HWU was set at the beginning and end of the simulation to simulate the actual harvest scenario.

2.4.2. Field-harvesting experiment method

To verify the effectiveness of the HUT collaborative operation strategy, a field-harvesting experiment was conducted at an unmanned rice farm at the Zengcheng Experimental Teaching Base of South China Agricultural University, Guangzhou, Guangdong, China.

To facilitate the unloading according to the estimated yield, the field to be harvested was chosen (Fig. 6). The longitude and latitude coordinates (WGS-84) of points A, B, C, and D were obtained according to the actual experimental requirements using DJI TERRA to form the boundary of the field to be harvested. The longitude and latitude coordinates of point U were obtained as the unloading points of the transporter to the truck. An east–north–up navigation plane coordinate system was used in this study [32]. Using point A as the origin of the navigation plane coordinate system, the longitude and latitude coordinates of points A, B, C, D, and U were converted into Gaussian projection plane coordinates and then into the navigation plane coordinate system. The coordinates of the key points are listed in Table 6. As shown in Table 6, the experimental field measured 5162.81 m².

The harvesting width was set to 1.9 m, and the path planning of the harvester was achieved using the parallel circling method. The final path plan included 1 round of a large-circle operation path and 14 rounds of nested operation paths (Fig. 7). The transporter paths included the HWU collaborative path H₁H₂, the TFU collaborative path T₁T₂, and guidance paths (Figs. 8(a) and (b)). Paths H₁H₂ and T₁T₂ were obtained using a translation method according to the path planning of the harvester and the required lateral spacing (set to 2.3 m) for collaborative unloading between two vehicles. Paths H₁H₂ and MU can be reused multiple times. However, path T₁T₂ must be planned and adjusted based on the real-time position of the harvester. The limited accuracy of the granary position sensors of the harvester and transporter led to inaccurate measurements and predictions of the overall granary information. As shown in Fig. 8(c), based on the yield estimation of the grain and the logical framework proposed in this study, paths for the transporter were designed. Invoking these paths and driving the transporter forward and backward allows collaborative grain unloading and transportation to be achieved. Each path of the harvester and transporter contains both location and status information related to the path, thus providing a reference for control. The definitions of the status-code information for the harvester and transporter are listed in Table 7, and the planning results are shown in Fig. 9.

The harvester and transporter speeds were set to 0.8 and 1.0 m∙s⁻¹, respectively. The path tracking of the harvester and transporter was performed using a preview-tracking control method [33]. The average absolute deviation and standard deviation of path tracking were 0.054 and 0.064 m, respectively. Communication between the harvester and transporter was achieved through radio and 4G technologies [34]. The data update frequency was stabilized at 10 Hz. The compensated predictor longitudinal control method was applied during the HWU collaborative operation for parking alignment [35], and the longitudinal accuracy and lateral alignment accuracy were less than 0.2 and 0.1 m, respectively. The gain self-adjusting single-neuron PID control method was used during the TFU collaborative operation to accurately align two vehicles [25], and the longitudinal alignment accuracy was less than 0.25 m. During the experiment, a communication link was maintained between the two vehicles, and the experimental data were monitored and recorded in real time.

2.4.3. Calculation of harvesting efficiency gains

The formula for calculating harvesting efficiency gains of the dual-unloading method compared with the single-harvester method and the only HWU mode is as follows:

(4)

\{\begin{matrix} η_{1} = \frac{t_{3}}{t_{1} + t_{2} + t_{3}} \\ η_{2} = \frac{t_{4}^{'}}{t_{1} + t_{2} + t_{4}} \end{matrix}

where t₁ is the time spent on harvesting, t₂ is the time spent on the harvester temporary parking, t₃ is the time spent on the harvester traverse to the edge of the field to unload, t₄ is the time spent on HWU, t₄′ is the parking time of HWU that TFU can save, η₁ is the harvesting efficiency gains compared with the single-harvester method, and η₂ is the harvesting efficiency gains compared with the only HWU mode.

3. Results

3.1. Simulation test

The simulation results are presented in Fig. 10. “State” indicates the status of the state actuator. Harvester states 1–7 represent A_h1, A_h2, A_h3-1, A_h3-2, A_h4, A_h5-1, and A_h5-2, respectively, whereas transporter states 1–9 represent A_t1-1, A_t1-2, A_t1-3, A_t2-1, A_t2-2, A_t2-3, A_t2-4, A_t3-1, and A_t3-2, respectively. The initial values of the states were set to 1, whereas the initial values of T_h1, T_h2, T_t1, T_t2, and T_t3 were set to 0. As shown in Fig. 10(a), the state actuators of the harvester and transporter can perform effective state transfers under the actions of the state evaluators, state verifications, and state triggers, thus accomplishing the HUT collaborative operation. Steps 2–5 and 43–46 relate to HWU collaborative unloading processes. Steps 12–15, 23–26, and 34–37 relate to TFU collaborative unloading processes. The triggering states are shown in Fig. 10(b). In summary, the state-transition chain and logical framework designed in this study are effective.

3.2. Field-harvesting experiment

During the harvesting period, TFU and HWU were performed six times and once, respectively, and the transporter unloaded grains into the truck four times. The harvester only stopped during HWU but continued to harvest at other times. The operational trajectories of the harvester and transporter are shown in Fig. 11 and Fig. 12 shows the experimental scenario of the HUT collaborative operation. During the collaborative grain-unloading process, the unloading barrel and harvester clutch effectively operated, and the grains were accurately unloaded into the granary of the transporter. The state actuators of the harvester and transporter were triggered through inter-vehicle communication. The triggering conditions for T_h1, T_h2, T_t1, T_t2, and T_t3 in the experiment are shown in Fig. 13, where “0” indicates an untriggered state, and “1” indicates that triggered state. The proposed logical framework of the HUT collaborative operation is effective and can be effectively combined with path planning, path-tracking control, inter-vehicle communication, collaborative operation control, and implementing control to realize independent harvesting, unloading, and transportation.

The duration of the field-harvesting application experiment was 73 min in total, which included 2 min for HWU performed once, and the harvest efficiency was 0.42 hm²∙h⁻¹. If the HUT collaborative operation is not adopted, then the harvester must traverse to the edge of the field to unload. Unloading was performed five times, and each unloading required 5 min, making the total unloading time 25 min. If only the HWU mode is adopted, then the collaborative unloading would require 10 min. Compared with the single-harvester unloading and only HWU, the HUT collaborative operation strategy combining both HWU and TFU increased the efficiency by 26.0% and 9.9%, respectively, according to Eq. (4). As the field area increased, the distance to the truck and the harvesting speed increased. The advantages of the HUT collaborative operation strategy are more prominent.

4. Discussion

Harvesting is an integral part of unmanned farms. Achieving autonomous harvesting, unloading, and transportation further improves harvesting efficiency for unmanned farms. In recent years, a number of researchers have studied collaborative operation methods and strategies. Most researchers are currently working on the design and optimization of path planning, scheduling methods, and task assignments for agricultural vehicles. Most studies have only performed simulation tests and given less consideration to practical applications. However, in the process of collaborative harvesting, there are strongly coupled stages between the harvester and transporter due to the precise collaborative unloading, making it insufficient to focus solely on path planning and scheduling methods. Therefore, this study proposed a HUT collaborative operation strategy using a dual-unloading mode by combining practical application scenarios with careful consideration of path planning, path-tracking control, inter-vehicle communication, collaborative operation control, implementation control, and other factors. Field-harvesting experiment revealed that the HUT collaborative operation strategy and cotransporter operation system proposed and designed in this study were able to achieve the full process of autonomous harvesting, unloading, and transportation, and its harvesting efficiency was improved by 26.0% compared with that of single-harvester. As shown in Table 8 [21], [36], [37], [38], [39], the HUT collaborative operation strategy has better practicability and adaptability than the collaborative operation method of agricultural machinery in harvesting application scenarios. It provides a new solution for collaborative operation.

Furthermore, the proposed HUT collaborative operation strategy must be combined with real-time and accurate path planning so that the transporter can further improve the efficiency of grain unloading and transportation. Real-time granary detection for the harvester and transporter is important for optimizing path planning, improving the efficiency of the transporter, and reducing fuel consumption. Because path planning needs to be predictive and collaborative, unloading cannot take place in unharvested areas. As a result, more information is needed from granary detection sensors. Therefore, the measurement range, accuracy, and stability of granary sensors must be improved, and models for granary and harvest quantities need to be developed. Future research should focus on combining accurate granary information to enable real-time dynamic path planning for the transporter, enhancing its efficiency and applicability.

5. Conclusions

This study provides a solution for autonomous HUT in the harvesting process of unmanned farms and evaluates the proposed HUT collaborative operation strategy through simulations and field experiments. The results and conclusions obtained are summarized as follows:

(1)The HUT strategy was developed by combining the two proposed unloading and transportation modes, HWU and TFU. The HUT collaborative operational logic framework was established by breaking down the primary stages of the collaborative operation, designing MSMs, and constructing state-transition chains. Subsequently, it was simulated and validated using MATLAB Simulink Stateflow. Simulation results showed that the designed state-transition chain and logical framework could effectively describe the entire process of HUT collaborative operation and achieve the state transfer of state actuators.

(2)Field-harvesting experiments were conducted. The field operation entailed a harvesting width of 1.9 m, a speed of 0.8 m∙s⁻¹, and a harvesting area of 5162.81 m². The collaborative harvesting system successfully completed the entire HUT operation. During the harvesting period, TFU was performed six times, HWU was performed once, and the transporter unloaded grains into the truck four times. The harvest efficiency was 0.42 hm²∙h⁻¹. The experimental results showed that the HUT collaborative operation strategy effectively integrated path planning, path-tracking control, inter-vehicle communication, collaborative operation control, and implementation control. Due to the significant reduction in parking time for unloading, harvesting efficiency increased by 26.0% and 9.9% compared with the single-harvester unloading and only HWU, respectively.

Hence, the HUT collaborative operation strategy and cotransporter system designed in this study can provide support for collaborative harvesting, beginning from theoretical models to technical implementation, and offer new solutions for the harvesting process of unmanned farms.

CRediT authorship contribution statement

Fan Ding: Writing – review & editing, Writing – original draft, Visualization, Validation, Software, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Xiwen Luo: Resources, Project administration, Conceptualization. Zhigang Zhang: Supervision, Resources, Funding acquisition, Conceptualization. Lian Hu: Validation, Supervision, Project administration, Funding acquisition. Xinluo Wu: Investigation, Data curation. Kaiyuan Bao: Software, Investigation. Jiarui Zhang: Investigation. Bingxuan Yuan: Investigation. Wenyu Zhang: Validation, Supervision, Resources, Funding acquisition, Conceptualization.

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.

Acknowledgments

This research was funded by the National Key Research and Development Program of China (2021YFD2000600), the National Natural Science Foundation of China (32071914), the Modern Agricultural Industry Technology System of China (CARS-170405), and the Key Research and Development Program (Science and Technology Demonstration Project) project of Shandong Province (2022SFGC0202). The authors gratefully acknowledge the editors and anonymous reviewers for their constructive comments on our manuscript. The authors also gratefully acknowledge the graduate students Minchang Wang, Hongkai Li, Haixiang Huang, Guocheng Zhang, Mingda Peng, Liwen Hu, and Tian Zhang for their experimental assistance in this paper.

Appendix A. Supplementary material

Supplementary data to this article can be found online at https://doi.org/10.1016/j.eng.2024.11.006.

References

Publishing order | Descend order by publishing year | Descend order by cited within

[1]	Li D, Li Z.System analysis and development prospect of unmanned farming.Trans Chin Soc Agric Mach 2020; 51(7):1-12.

[2]	Luo X, Liao J, Hu L, Zhou Z, Zhang Z, Zang Y, et al.Research progress of intelligent agricultural machinery and practice of unmanned farm in China.J South China Agric Univ 2021; 42(6):8-17.

[3]	Zhang Z, Luo X, Zhao Z, Huang P.Trajectory tracking control method based on kalman filter and pure pursuit model for agricultural vehicle.Trans Chin Soc Agric Mach 2009; 40(S1):6-12.

[4]	Chou H, Khorsandi F, Vougioukas S, Fathallah F.Developing and evaluating an autonomous agricultural all-terrain vehicle for field experimental rollover simulations.Comput Electron Agric 2022; 194:106735.

[5]	Sun H, Slaughter D, Ruiz M, Gliever C, Upadhyaya S, Smith R.RTK GPS mapping of transplanted row crops.Comput Electron Agric 2010; 71(1):32-37.

[6]	Ding Y, Xia Z, Peng J, Hu Z.Design and experiment of the single-neuron PID navigation controller for a combine harvester.Trans Chin Soc Agric Eng 2020; 36(7):34-42.

[7]	Xue J, Zhang L, Grift T.Variable field-of-view machine vision based row guidance of an agricultural robot.Comput Electron Agric 2012; 84:85-91.

[8]	Zhang Y, Li Y, Liu X, Tao J, Liu C, Li R.Fuzzy adaptive control method for autonomous rice seeder.Trans Chin Soc Agric Mach 2018; 49(10):30-37.

[9]	Niu X, Gao G, Bao Z, Zhou H.Path tracking of mobile robots for greenhouse spraying controlled by sliding mode variable structure.Trans Chin Soc Agric Eng 2013; 29(02):9-16.

[10]	Xu G, Chen M, He X, Pang H, Miao H, Cui P, et al.Path following control of tractor with an electro–hydraulic coupling steering system: layered multi-loop robust control architecture.Biosyst Eng 2021; 209:282-299.

[11]	He J, Hu L, Wang P, Liu Y, Man Z, Tu T, et al.Path tracking control method and performance test based on agricultural machinery pose correction.Comput Electron Agric 2022; 200:107185.

[12]	Chi R, Xiong Z, Jiang L, Ma Y, Huang X, Zhu X.Path tracking control algorithm of transplanter based on model prediction.Trans Chin Soc Agric Mach 2022; 53(11):22-30.

[13]	Zhang X, Geimer M, Noack P, Grandl L.A semi-autonomous tractor in an intelligent master–slave vehicle system.Intell Serv Robot 2010; 3(4):263-269.

[14]	Li S, Cao R, Wei S, Ji Y, Zhang M, Li H.Development of multivehicle cooperative navigation communication system based on TD-LTE.Trans Chin Soc Agric Mach 2017; 48(S1):45-51.

[15]	Zhang C, Noguchi N, Yang L.Leader–follower system using two robot tractors to improve work efficiency.Comput Electron Agric 2016; 121:269-281.

[16]	Chen J, Fu S, Guan Z, Zhu F, Zhu L, Xia H, et al.Communication method for combine harvester group using Lora technology.Trans Chin Soc Agric Eng 2022; 38(16):81-89.

[17]	Yao J, Teng G, Huo L, Yuan Y, Zhang F.Optimization of cooperative operation path for multiple combine harvesters without conflict.Trans Chin Soc Agric Eng 2019; 35(17):12-18.

[18]	Zhai Z, Wang X, Wang L, Zhu Z, Du Y, Mao E.Collaborative path planning for autonomous agricultural machinery of master slave cooperation.Trans Chin Soc Agric Mach 2021; 52(S1):542-547.

[19]	Zhang F, Luo X, Zhang Z, He J, Zhang W.Agricultural machinery scheduling optimization method based on improved multi-parents genetic algorithm.Trans Chin Soc Agric Eng 2021; 37(9):192-198.

[20]	Wang N, Yang X, Wang T, Xiao J, Zhang M, Wang H, et al.Collaborative path planning and task allocation for multiple agricultural machines.Comput Electron Agric 2023; 213:108218.

[21]	Noguchi N, Will J, Reid J, Zhang Q.Development of a master–slave robot system for farm operations.Comput Electron Agric 2004; 44(1):1-19.

[22]	Bai X, Wang Z, Hu J, Gao L, Xiong F.Harvester group corporative navigation method based on leader–follower structure.Trans Chin Soc Agric Mach 2017; 48(7):14-21.

[23]	Li S, Xu H, Ji Y, Cao R, Zhang M, Li H.Development of a following agricultural machinery automatic navigation system.Comput Electron Agric 2019; 158:335-344.

[24]	Zhang W, Zhang Z, Luo X, He J, Hu L, Yue B.Position–velocity coupling control method and experiments for longitudinal relative position of harvester and grain truck.Trans Chin Soc Agric Eng 2021; 37(9):1-11.

[25]	Ding F, Zhang W, Luo X, Hu L, Zhang Z, Wang M, et al.Gain self-adjusting single neuron pid control method and experiments for longitudinal relative position of harvester and transport vehicle.Comput Electron Agric 2023; 213:108215.

[26]	Cho H, Hachtel GD, Macii E, Plessier B, Somenzi F.Algorithms for approximate fsm traversal based on state space decomposition.IEEE Trans Comput Aided Des Integrated Circ Syst 1996; 15(12):1465-1478.

[27]	Leonard J, How J, Teller S, Berger M, Campbell S, Fiore G, et al.A perception-driven autonomous urban vehicle.J Field Robot 2008; 25(10):727-774.

[28]	Talebpour A, Mahmassani H, Hamdar S.Modeling lane-changing behavior in a connected environment: a game theory approach.Transp Res Procedia 2015; 7:420-440.

[29]	Xiong L, Jia T, Chen J, Xing X, Li B.Hazard identification method for safety of the intended functionality based on finite state machine.J Tongji Univ 2023; 51(4):616-622.

[30]	Yang L, Tang X, Wu S, Wen L, Yang W, Wu C.Local path planning for autonomous agricultural machinery on farm road.Trans Chin Soc Agric Eng 2024; 40(01):27-36.

[31]	Fue K, Porter W, Barnes E, Li C, Rains G.Center-articulated hydrostatic cotton harvesting rover using visual-servoing control and a finite state machine.Electronics 2020; 9(8):1226.

[32]	He J, Luo X, Zhang Z, Wang P, He J, Yue B, et al.Positioning correction method for rice transplanters based on the attitude of the implement.Comput Electron Agric 2020; 176:105598.

[33]	Wang H, Wang G, Luo X, Zhang Z, Gao Y, He J, et al.Path tracking control method of agricultural machine navigation based on aiming pursuit model.Trans Chin Soc Agric Eng 2019; 35(4):11-19.

[34]	Ding F, Zhang W, Luo X, Zhang Z, Wang M, Li H, et al.Design and experiment for inter-vehicle communication based on dead-reckoning and delay compensation in a cooperative harvester and transport system.Agriculture 2022; 12(12):2052.

[35]	Zhang W, Hu L, Ding F, Luo X, Zhang Z, Hu L, et al.Parking precise alignment control and cotransporter system for rice harvester and transporter.Comput Electron Agric 2023; 215:108443.

[36]	Tian Y, Bhattacharya S.Smart autonomous grain carts for harvesting-on-demand.In: IEE E/RSJ International Conference on Intelligent Robots and Systems (IRO S); 2017 Dec 24–28; Vancouver, B C, Canada. Piscataway: IEE E; 2017. p. 5168–73.

[37]	He P, Li J.The two-echelon multi-trip vehicle routing problem with dynamic satellites for crop harvesting and transportation.Appl Soft Comput 2019; 77:387-398.

[38]	Zhang W, Zhang Z, Zhang F, Ding F, Hu L, Luo X.Cooperative autonomous operation strategy and experiment of the rice harvester together with a rice-transporting vehicle.Trans Chin Soc Agric Eng 2022; 38(15):1-9.

[39]	Li S, Zhang M, Wang N, Cao R, Zhang Z, Ji Y, et al.Intelligent scheduling method for multi-machine cooperative operation based on NSGA-III and improved ant colony algorithm.Comput Electron Agric 2023; 204:107532.

RIGHTS & PERMISSIONS

THE AUTHOR