Soil Moisture Detection and Linear Deceleration Control Strategy Enhancing Trenching Depth Precision and Stability for Rapeseed Sowing

Abstract

Sowing depth significantly affects the germination of rapeseed, and different soil moisture conditions require corresponding sowing depths. However, most current trenching devices do not account for soil moisture content, and commonly used hydraulic or constant-force trenching equipment also exhibits deficiencies in stability and consistency. To address these challenges, this study developed an automatic depth adjustment control system based on soil moisture content. A soil moisture detection device and an innovative sliding mechanism that maintained the soil moisture sensor in a relatively stationary position relative to the soil during seeder movement were introduced. An automatic sowing depth adjustment device was designed to modulate the sowing depth. A control strategy that incorporated the Kalman filtering algorithm and linear deceleration equations was conducted. At an observation noise covariance matrix (Q/R) of 0.001, a deceleration range of 40 mm and a minimum speed of 10, the control system exhibited minimal overshoot (approximately 4%) and steady-state error (approximately 3.2 mm). It effectively adjusted the trenching depth while operating at speeds ranging from 2 to 3.6 km/h, successfully adapting to variations in soil topography. The system performance tests revealed that the control system adjustment time (t_s) was 534 ms and the steady-state error remained within 1 mm. Under three different soil moisture content conditions, the sowing depth qualification rate and stability coefficients consistently surpassed 90% and 80%, respectively. This research offers a sowing depth adjustment control system based on soil moisture content, contributing to more precise depth regulation for rapeseed sowing.

1. Introduction

Rapeseed is a significant oil crop in China, and enhancing the quality of seed sowing is a critical approach to ensuring the high-quality growth for rapeseed. An appropriate sowing depth can provide favorable conditions for the germination of rapeseed, which is determined by the furrow system of seeder [1]. An automatic furrow depth adjustment system enhances a seeder’s adaptability to varied field conditions, improving seeding precision and serving precision agriculture more effectively [2]. In most seeding machines, depth adjustment mechanisms utilize pressure sensors to detect deviations and, in conjunction with established furrow depth mathematical models, make adjustments, thereby improving the stability and uniformity of seeding depth. The prevalent method is fixed-depth adjustment, emphasizing consistency, demonstrating robust adaptability under complex field conditions, and garnering positive feedback in numerous agricultural practices.

In actual agricultural production, the conditions of farmland across various regions exhibit significant spatial variations in factors such as soil moisture content and compactness. This complexity makes precise control of sowing depth during the sowing process more challenging [3,4]. Its adaptation algorithms for complex fields include Kalman filter fusion algorithm [5], Mamdani fuzzy algorithm [6], PID fuzzy algorithm [7], etc. Different algorithms are combined with appropriate sensors (commonly used sensors include Piezoelectric thin film sensor [8], laser triangulation displacement sensors [9], pressure sensors [10,11], angle sensors, array radar sensors, etc.) to establish corresponding intelligent adjustment models [12,13], optimizing the control of seeding depth [4,5,6,7,8,9,10,11,12,13,14,15,16] to reduce seeding errors and achieve real-time adaptive fixed depth adjustment with uniform seeding.

Recent research on fixed-depth adjustment has combined the Mamdani and PID algorithms with a sowing depth detection system to achieve uniform real-time adaptive adjustment of sowing depth [17]. Despite these advancements, there is still room for improvement due to minor fluctuations in sowing depth. Efficient sowing depth control should provide accurate decisions based on different sowing methods and mechanical structures, adjusting and controlling sowing depth precisely through driving mechanisms [18,19]. Studies have shown that, in high-speed operations, the parallel four-bar linkage mechanism (PFPM) and spring-assisted downward pressure (SAD) with an upward tilt improve the stability of ground wheel downward pressure (GWD) and the sowing depth of no-till planter units [20], offering theoretical support for the design and application adjustment of high-speed no-till planters. Additionally, structured light cameras are used to monitor the trenching depth of planters in real time and adjust soil coverage and compaction force accordingly [21].

The sowing depth of winter rape is typically regulated to be within 2 cm, which differs from other crops. [22]. However, local water-drought rotation methods [23], heavy soil, and concentrated autumn rain during the seedling stage can cause significant humidity differences in the field [24], leading to severe waterlogging [25]. From 2022 to 2023, the research team employed fixed-depth adjustment methods and proposed “stubble connection + trenching drainage + simultaneous fertilization” technology to address issues encountered during the rapeseed sowing period in Guang’an District, Guang’an City. This approach significantly improved the mechanization rate and seedling emergence rate.

As research progresses, narrowing the observation scale to the same field with varying moisture content in different areas reveals that using fixed-depth adjustment methods for sowing may not guarantee seedling emergence rates in areas with higher humidity [26,27,28], and replanting may still be needed later to ensure the emergence rate.

Although the adaptation algorithms have been conducted for fixed depth adjustment, research on dynamic seeding depth adjustment is still limited. The objective of this research was to: (1) propose the design and experiment of an automatic adjustment system for rapeseed seeding depth adaptive to soil moisture content; (2) test the application of the Kalman filter algorithm in conjunction with the linear deceleration equation to regulate the key components involved in furrow opening, thereby enhancing the dynamic stability of furrow depth in relation to soil moisture content. Experimental tests determined the optimal control parameters, enhancing the dynamic stability of the seeding furrow depth and its synchronization with soil moisture content.

2. Materials and Methods

2.1. Planting Pattern of Rape Seeding

The mechanical seeding of rapeseed conventionally involves sowing 2 to 3 seeds per hole at an appropriate depth to ensure optimal conditions for seed emergence and seedling growth [29,30]. Soil moisture content significantly affects the seedling growth and yield for rapeseed. According to agronomic requirements, when soil moisture content is below 20%, a seeding depth of approximately 20 mm was recommended to maintain soil moisture favorable for germination. For moisture content between 20% and 35%, a seeding depth of about 10 mm was advised. In instances where soil moisture content exceeded 35%, seeds should be placed on the soil surface to prevent waterlogging damage [22] (Figure 1).

2.2. The Structure and Working Principle of the Seeder

The soil moisture-based rapeseed seeding depth adjustment system comprises a soil moisture content detection device and an automatic seeding depth adjustment mechanism (Figure 2). The soil moisture content detection device was situated on the lateral aspect of the seeder, while the depth adjustment mechanism was located at the posterior terminus of the rapeseed seeder. The depth measurement device was affixed to the seeding depth adjustment mechanism, aligning with the furrow opener. The depth measurement device was positioned on the front side of the furrow opener’s installation location. In operational mode, the soil moisture-based rapeseed seeding depth adjustment mode was initially selected through the human-computer interaction interface (9), the furrow discrepancy threshold was designated, and subsequently the procedure was activated by depressing the initiation button. The soil moisture content detection device evaluated the moisture content across diverse points, thereby signaling the control system to establish the intended seeding depth for said point. Upon integrating the data with measurements from the depth gauge, the pushing rod motor (4) was compelled to make the necessary adjustments to the seeding depth. This ensured that the furrow opener dynamically adapted to varying moisture content conditions, guaranteeing an optimal seeding depth.

2.3. Hardware Design

2.3.1. Soil Moisture Detection Device

The soil moisture detection device, mounted on the chassis, integrated components such as soil moisture content sensors (Dalian Zheqin Technology Enterprise Company Limited, Dalian, China), a re-insertion soil mechanism, a step motor (Wenzhou Lucheng District Xingkong Electrical Equipment Company Limited, Wenzhou, China), a constant force spring assembly, a linear guide track, and a wire guide slider (Figure 3). The re-insertion mechanism featured a disc, an annular guide rod, and an annular groove. Notably, the soil moisture content sensor was affixed to this ring-shaped guide rod. Actuated by the motor, the disc executed rotational motion, which propelled the guide rod vertically, thereby facilitating the height adjustment of the soil moisture content sensor (Figure 4). Stringent control over the disc’s rotation angle and cycle ensures that each revolution corresponded to a semicircular path, enabling automated insertion and extraction of the sensor from the soil.

The constant force spring mechanism, secured to the terminus of an aluminum extrusion, anchored the constant force spring to the right flank of the soil insertion assembly. This assembly was mounted on a linear guide rail that extends 120 cm in length and was elevated 22 cm above ground level. The constant force spring was engineered to exert a consistent force of 1.225 N. Upon data acquisition for soil moisture content and subsequent soil removal, the soil insertion assembly was restored to its initial configuration through the action of the springs. Within the confines of the soil insertion apparatus, the disc’s upper limit stood at 31 cm, while its lower bound was set at 25 cm, yielding a penetrative depth of 6 cm into the soil.

2.3.2. Automatic Sowing Depth Adjustment Device

The automatic adjustment mechanism for sowing depth consisted of a depth adjustment apparatus, a human-computer interface, an alarm system, a depth detection sensor, and a speed sensor (Maibaolai Electronic Technology Company Limited, Suzhou, China). For depth measurement, laser distance sensors (model L1s) (Shenzhen Motian RF Technology Company Limited, Guangzhou, China). are employed in this study due to their superior accuracy, rapid response, and minimal environmental influence. The depth adjustment apparatus comprised a disc furrow opener, a pushing rod motor (Changzhou Luzhiren Technology Company Limited, Guangzhou, China), and a furrowing lifting frame. To meet the furrowing speed requirements, a pushing rod motor. drove the furrowing lifting frame. The furrowing lifting frame, with a parallelogram hinged four-bar structure, allowed the pushing rod motor to control the vertical movement of the furrow opener, thereby adjusting sowing depth.

The human-computer interface employed a TFT LCD display (Guangzhou Dacai Optoelectronics Technology Company Limited, Guangzhou, China), with interfaces developed using Visual TFT software (Version 3.0.0.1243). This included a parameter setup interface (Figure 5a) and an abnormal situation response interface (Figure 5b). The functionalities of the interface comprised threshold setting for ditching, pop-up alarms, and real-time feedback of distance measurement data.

2.4. Control Strategy

2.4.1. Control Method of Soil Moisture Detection System

The Arduino Mega 2560 (Arduinor Company Limited, Chiasso, Switzerland) microcontroller generated pulses that controlled the TB6600S driver, with a subdivision ratio set at 6400. The microcontroller emitted 3200 pulses to rotate the stepper motor by a semicircular path, which in turn activated the cam mechanism to insert and subsequently extract the sensor from the soil. Figure 6a provides a visual representation of the initial position of the moisture content detection system, where the device was protected from horizontal external forces and positioned adjacent to the constant force spring structure. The sensor was positioned above the soil surface, with its ring groove facing upwards. (1) Upon activation, the system initiated a rotation of the stepper motor by half a circle, resulting in the disc cam’s rotation. This mechanical action forced the annular guide rod downwards by the annular groove, thereby embedding the sensor into the soil as shown in Figure 6b. (2) The sensor was maintained within the soil for a duration of 1 s to ensure the recording of stable data. Concurrently, the seed drill is moving forward. In order to reduce the damage of the sensor in the soil, the sensor remains stationary relative to the soil and moves on the slide rail. Due to the seed drill moving forward and the horizontal resistance exerted by the soil, the moisture detection device pulls the spring away from the constant force spring structure and the sensor slides on the rail, as demonstrated in Figure 6c. (3) Following the data collection phase, the stepper motor executed another half-circle rotation, facilitating the extraction of the sensor from the soil. Pulled by the constant force spring, the moisture detection device relocated closer to the constant force spring structure on the slide rail, ultimately returning to its origination point. This iterative process enabled the measurement of soil moisture at diverse locations.

2.4.2. Control Strategy of Automatic Adjustment System for Sowing Depth

(1)

Kalman filter algorithm.

In field operations, the data collected by sensors can be significantly influenced by environmental noise and mechanical vibrations. To address this challenge, Kalman filtering was utilized to reduce the adverse effects of field noise and mechanical vibrations on sensor data accuracy. The methodology for implementing the Kalman filter algorithm involves the following steps [31]:

$$\mathrm{a}=1, \left\{\begin{array}{l}\widehat{X_0}=\left[D_t\right]\\ P_0=\left[\begin{array}{cc}{\sigma }^2& 0\\ 0& {\sigma }^2\end{array}\right]\end{array}\right.$$

(1)

where, $\widehat{X_0}$ is the initial state information of the system; D_t is the initial distance measurement value of the laser sensor, mm (D_t = 0); P₀ is to the covariance matrix; and σ is the initial covariance matrix coefficient (σ = 1 in this research).

Prediction of system state and covariance matrix were as follows:

$$\left\{\begin{array}{l}\widehat{X_t}=\widehat{F{X}_{t-1}}+B{U}_t+W\\ {P_t}^{-}=F{P}_{t-1}{F}^T+Q\end{array}\right.$$

(2)

where, F is the state transfer matrix (F = 1 in this study); $F^T$ is the transpose of the state transfer matrix; $\widehat{X_t}$ is the state estimate at time t; $\widehat{X_{t-1}}$ is the state estimate at time t − 1; B is the control matrix (B = 0 in this study); $U_t$ is the control quantity; W is the distance estimation noise; $P_t^{-}$ is the covariance matrix of the prior estimation error; $P_{t-1}$ is the covariance matrix of the prior estimation error; and Q is the state transfer covariance matrix.

Next, the Kalman gain was calculated.

$$K_t={\displaystyle \frac{P_t^{-}{H}^T}{H{P}_t^{-}{H}^T+R}}$$

(3)

where, $K_t$ is the Kalman gain matrix at time t and H is the observation matrix (H = 1 in this study). $H^T$ is the transpose of the observation matrix; R is the covariance matrix of the observation noise.

The measurement estimate was calculated.

$$Z_t=H{X}_t+V$$

(4)

where, $Z_t$ is the distance observation value at time t; V is the observation value noise; and $X_t$ is the measured value of the distance at time t.

Hence, the system state and covariance matrix are obtained as follows:

$$\left\{\begin{array}{l}{P}_t=(I-K_{t}H)P_t^{-}\\ \widehat{X_t}=\widehat{X_t^{-}}+K_t(Z_t-H\widehat{X_t^{-}})\end{array}\right.$$

(5)

where, I is the unit matrix; $\widehat{X_t^{-}}$ is the prior state estimate at time t; and $P_t$ is the a posteriori estimate error covariance matrix. By synthesizing Equations (2) and (3),

$$K_t={\displaystyle \frac{\frac{P_{t-1}}{R}+\frac{Q}{R}}{\frac{P_{t-1}}{R}+\frac{Q}{R}+1}}$$

(6)

From Equation (6) and simulation results using Python (Version 3.8.9), it was evident that the filter’s performance was contingent on the Q/R ratio. The subsequent discussion delved into the nuanced influence of Q/R on filter efficacy. A judicious approach to modulating Q and R involves selecting an appropriate Q value while introducing adequate uncertainty into the process signal. This facilitated the development of a rudimentary process model capable of yielding satisfactory results [30]. Simulations revealed that a combination of Q = 0.001 and R = 1 resulted in data that closely aligned with the true values, exhibiting minimal fluctuations. Consequently, the optimal filter parameters for this system were determined to be Q = 0.001 and R = 1, as depicted in Figure 7.

(2)

Mathematical model of control system.

Figure 8 depicts a schematic of the automatic sowing depth adjustment system’s key components, which are comprised of a parallelogram hinged four-bar mechanism (ABCD), a pushing rod motor (CE), a laser distance sensor, and a double disc furrow opener. The control methodology for the automatic sowing depth adjustment system is as follows:

(3)

Delay calculation strategy.

To prevent measurement interference from the furrow opener, the laser sensor was positioned at the front of its installation point. This setup introduced a delay between the sensor’s measurement and the depth adjustment mechanism’s response, potentially causing inaccuracies. To rectify this, a delay compensation technique was utilized. The double-disc furrow opener’s projection on the mid-plane between the discs forms an ellipse. The lateral relationship between the laser sensor and the furrow opener is illustrated in Figure 9.

The delay in the depth adjustment mechanism’s response, alongside the trenching depth and the relative positions of the laser sensor and trencher, are as follows:

$$\left\{\begin{array}{l}{\displaystyle \frac{y^2}{a^2}}+{\displaystyle \frac{x^2}{b^2}}=1\\ y=-a+h_0\\ t^{\prime }={\displaystyle \frac{m+k}{v^{\prime }}}\\ a={\displaystyle \frac{D}{2}}\\ b=D\times \cos {\displaystyle \frac{\alpha }{2}}\times {\displaystyle \frac{1}{2}}\end{array}\right.$$

(7)

where, h₀ is the furrowing depth, mm; y is the ordinate of the contact position between the soil and the furrow opener when the furrowing depth is h₀, mm; x is the abscissa of the contact position between the soil and the furrow opener when the furrowing depth is h₀, mm; a is the major axis of the ellipse, mm; b is the minor axis of the ellipse, mm; m is the horizontal distance between the laser ranging sensor and the front end of the disc furrow opener, mm; k is the horizontal distance between the contact position between the soil and the furrow opener and the front end of the furrow opener when the furrowing depth is h₀, mm; D is the diameter of the furrow opener; $v^{\prime }$ is the driving speed of the rapeseed sowing robot, m/s; α is the angle between the two disks of the double disc furrow opener.

Therefore, the action lag time of the depth adjustment device is:

$$t^{\prime }={\displaystyle \frac{\cos \frac{\alpha }{2}\left(\frac{D}{2}-\sqrt{D{h}_0-{h_0}^2}\right)+m}{v^{\prime }}}$$

(8)

Figure 10 illustrates the method used to determine the velocity of the sowing robot, which employs a rotary encoder. The detection of 1000 pulses by the single-chip microcomputer indicates that the seeder has covered a specific distance, denoted as l. The duration required for this movement is t, and, subsequently, the speed can be calculated as $v\prime ={\displaystyle \raisebox{1ex}{$l$}\!\left/ \!\raisebox{-1ex}{$t$}\right.}$.

(4)

Control of actuator motor.

The system employed a linear deceleration equation for control purposes, establishing a linear correlation between error distance and speed range. This approach allowed the determination of pushing rod velocities at various errors, facilitating variable speed regulation. The specific pushing rod motor speed and PWM value are as follows:

$$V={\displaystyle \frac{(V_{\max }-V_{\min })}{R_d}}\cdot E_d+V_{\min }$$

(9)

$$PWM={\displaystyle \frac{V\cdot 255}{100}}$$

(10)

where, V is an integer value between 0 and 100, representing the speed of the actuator motor, and has no unit; $E_d$ is the error distance, mm; $R_d$ is the deceleration interval, mm; $V_{max}$ indicates that the actuator motor runs at full speed, takes 100, and has no unit; and $V_{\min }$ indicates that the actuator motor runs at the minimum speed, and has no unit. $R_d$ and $V_{min}$ determine the control system performance.

The system unit step response curve, influenced by varying control parameters, is presented in Figure 11. With a deceleration range of 10 mm, the furrow opener exhibited a notably brief deceleration buffer duration, leading to pronounced oscillations within the control system (Figure 11a). Conversely, when the deceleration range was set to 50 mm, the furrow opener functioned at a reduced pace, resulting in an excessively prolonged system adjustment period (Figure 11b). The optimal deceleration buffer duration was 40 mm, during which the minimum velocity attained was 10.

The system demonstrated a constant error of 2 mm under optimal control parameters, which include a deceleration range of 40 mm and a minimum speed of 10 (Figure 12). Within the specified error threshold, the static response time was measured to be 1323 ms, indicating a relatively short response time for the system. The effect comparison data under each control parameter are shown in

Table 1. Comparison of effects under each control parameter.

Parameter		Adjusting Time (ms)	Steady State Error (mm)
Deceleration Interval Rd (mm)	Rod Motor Speed V
25	0	3390	1
	25	1823	1
	50	-	-
30	0	2476	1
	25	2002	0
	50	3034	2
35	0	1655	3
	25	1537	4
	50	2482	4
40	10	1323	2
45	0	3564	1
	25	2421	1
	50	3910	2
50	0	26,049	5
	25	5458	1
	50	9194	2

.

2.4.3. Control Method for Adjusting Sowing Depth Based on Soil Moisture Content

The sowing depth adjustment system, illustrated in Figure 13 with the user setting an initiation, is triggered by pressing the start button, prompting the system into action. (1) In operation, the rotary encoder detected the seeder’s velocity to determine the delay control duration for furrowing. Concurrently, soil moisture was gauged by a dedicated detection device. (2) The Arduino microcontroller ascertained sowing depth based on these moisture data. If Laser ranging sensor data are deemed normal, they undergo Kalman filtering to mitigate measurement inaccuracies arising from mechanical vibrations. (3) Subsequently, the furrowing lifting frame height corresponding to the sowing depth was deducted from the filtered measurement, yielding the necessary adjustment distance for the frame. If this exceeded the error threshold, a delayed activation of the pushing rod adjusted the furrowing lifting frame. Conversely, if it remained beneath the threshold, signifying the soil depth was within an acceptable margin of error, no adjustment is executed. Abnormal ranging, indicated by a sensor data value of 0, activated a flashing buzzer alarm and presented an alert via a pop-up window on the display, notifying the user.

3. Results and Analysis

3.1. Experimental Test Conditions and Equipment

To validate the precision and reliability of the developed system, a test was conducted in the soil bin laboratory at Sichuan Agricultural University. The instruments mainly included a soil trough test bench, a rapeseed sowing seeder, a PC monitoring terminal, a tape measure, and a spiral micrometer (Figure 14). The performance test of the rapeseed sowing seeder conformed to the directives of NT/T740-2003. During the test, the system’s control performance, sowing depth index, and trenching capability under soil undulation were scrutinized (Equation (11)). Additionally, a comparative analysis was undertaken between automatic and non-automatic depth adjustment mechanisms.

$$\left\{\begin{array}{l}\stackrel{-}{h}={\displaystyle \frac{{\displaystyle \sum _{i=1}^q{h}_i}}{Q}}\\ \sigma =\sqrt{{\displaystyle \frac{{\displaystyle \sum _{i=1}^q(h_i-\stackrel{-}{h}{)}^2}}{Q}}}\\ V={\displaystyle \frac{\sigma }{\stackrel{-}{h}}}\times 100\%\\ U=1-V\end{array}\right.$$

(11)

where, $\overline{h}$ represents the mean sowing depth, mm; h_i denotes the specific sowing depth, mm; σ is the standard deviation of the sowing depth, mm; V signifies the coefficient of variation for the sowing depth as a percentage, %; and U indicates the stability coefficient of the sowing depth as a percentage, %.

3.2. Control Performance Test

Figure 15 illustrates the control system’s output waveform when a target trenching depth of 10 mm was specified. In response to the pushing rod motor, the furrow opener descended. A measurement delay (t_d) of 142 ms recorded a trenching depth of 5 mm. Soil resistance induced fluctuations in the furrow opener’s trajectory, which were mirrored by variations in the depth curve. By t = 400 ms, the trenching depth first achieved the desired 10 mm. Inertia carried the opener beyond this point, leading to an overshoot that peaked at 11 mm. After systemic oscillations, the trenching depth stabilized at 11 mm by an adjustment time (t_s) of 534 ms. This final trenching depth of 11 mm was within the acceptable error margin of 2 mm. The system exhibited a responsive adjustment time of 534 ms. Figure 16 provides a scan of the furrowing results, demonstrating consistent depth and linear trajectory.

3.3. Sowing Depth Accuracy Test

To verify the system’s control accuracy, three series of tests were performed under different soil moisture content levels. Each series encompassed two trenching strokes, with 30 points sampled at consistent intervals along the stroke’s trajectory. The trenching depth was assessed manually thrice per point using a micrometer.

During the experiment, a permissible error margin of 2 mm for trenching depth was set.

Table 2. Sowing depth under three soil moisture content conditions.

Test Number	Actual Sowing Depth (mm)			Absolute Error of Seeding Depth (mm)
	Soil Moisture Content < 20%	20% < Soil Moisture Content < 35%	Soil Moisture Content > 35%
1	18.2	8.6	1.5	1.57 ± 0.0785
2	18.9	9.1	1.1	1.03 ± 0.0515
3	18.1	10.8	1.2	1.30 ± 0.065
4	19.2	11.2	0.9	0.97 ± 0.0485
5	18.3	11.3	1.3	1.43 ± 0.0715
6	19.1	10.9	1.4	1.07 ± 0.0535
7	21.7	11.2	1.2	1.37 ± 0.0685
Average	19.07	10.44	1.24	1.25 ± 0.0625

illustrates that the average absolute error for trenching depth across three soil moisture content levels varied from a minimum of 0.97 mm to a maximum of 1.57 mm. The overall mean absolute error was recorded at 1.25 mm, which was within the prescribed error limit of 2 mm. The sowing depth qualified rate and stability coefficients for sowing depth consistently exceeded 90% and 80%, respectively (Figure 17). The coefficient of variation of sowing depth increased with an increase in soil moisture content, aligning with the predefined standards.

3.4. Depth Automatic Adjustment and Non-Adjustment Comparison Test

Table 3. Comparison of sowing between automatic and non-adjustable treatments.

Index	No Adjustment	Automatic Adjustment
Sowing depth qualified rate (%)	30.8	92.5
Coefficient of variation of seeding depth (%)	56.46	10.65
Seeding depth stability coefficient (%)	43.54	89.35
Standard deviation of seeding depth (mm)	3.41	1.11

illustrates the sowing depth differences between automatic adjustment and non-adjustment. The qualification rate of sowing depth with automatic adjustment increased by 61.7% points compared to the non-adjusted treatment. The coefficient of variation decreased by 45.81 percentage points, and the standard deviation reduced by 2.30 mm, indicating a significant improvement in furrowing precision and uniformity. Figure 18a shows that the sowing depths under automatic adjustment are within the error range, closely aligning with the target sowing depth and exhibiting minimal fluctuations and uniform distribution. In contrast, most sowing depths without adjustment fall outside the error range. When the soil undulation height was approximately 5 cm, the sowing depth mirrored the soil fluctuations in instances where furrowing was not automated (Figure 18b). However, upon integration of the automatic control system, the sowing depth was consistently maintained around the intended value of 20 mm. This indicated the automatic adjustment sowing depth control advantage.

The aforementioned soil trough test was conducted to ascertain the evaluation indices of the control system, whereas field tests provided a more direct representation of the system’s practical efficacy. Field test results revealed that the emergence rates of rapeseed were 63%, 60%, and 58% for soil moisture contents below 20%, between 20% and 35%, and above 35%, respectively. These findings underscored the exceptional performance of our system in actual agricultural settings.

4. Conclusions

This study presents a dynamic seeding depth adjustment system for rapeseed based on soil moisture detection. The soil moisture detection device and automatic sowing depth adjustment device were designed. The control performance and sowing depth tests were conducted to verify the control system. The main contributions of this study can be summarized as follows.

(1)

The soil moisture detection device was designed to automating test the soil moisture content sensor. A control equation of “linear deceleration” and a Kalman filter algorithm were proposed to ensure precise operation of the furrowing depth. The optimal parameters were an observation noise covariance matrix (Q/R) of 0.001, a deceleration range of 40 mm, and a minimum speed of 10. This control strategy therefore shortened response times and enhanced stability and precision.

(2)

The simulation and test of the control system’s precision and reliability were conducted. The control system adjustment time (t_s) was found to be 534 ms and the steady-state error was within 1 mm.

(3)

The trenching depth test demonstrated that under three different soil moisture content conditions, the sowing depth qualification rate and stability coefficients consistently surpassed 90 and 80%, respectively. The qualification rate of sowing depth with automatic adjustment increased by 61.7 percentage points compared to the non-adjusted treatment. The coefficient of variation decreased by 45.81 percentage points. This suggested that the automatic adjustment sowing depth system exhibited excellent performance, characterized by high control accuracy and good stability. Consequently, this control system can more effectively regulate sowing depth.

The limitation of this research lies in the need to enhance the response speed of the control system to improve both the system’s operational efficiency and the precision of seeding depth. Future endeavors will focus on optimizing this system through advanced control methodologies.