Design of the Mechanical Structure of a Field-Based Crop Phenotyping Platform and Tests of the Platform

Abstract

The field mobile platform is an important tool for high-throughput phenotype monitoring. To overcome problems in existing field-based crop phenotyping platforms, including limited application scope and low stability, a rolling adjustment method for the wheel tread was proposed. A self-propelled three-wheeled field-based crop phenotyping platform with variable wheel tread and height above ground was developed, which enabled phenotypic information of different dry crops in different development stages. A three-dimensional model of the platform was established using Pro/E; ANSYS and ADAMS were used for static and dynamic performance. Results show that when running on flat ground, the platform has a vibration acceleration lower than 0.5 m/s². When climbing over an obstacle with a height of 100 mm, the vibration amplitude of the platform is 88.7 mm. The climbing angle is not less than 15°. Field tests imply that the normalized difference vegetation index (NDVI) and the ratio vegetation index (RVI) of a canopy measured using crop growth sensors mounted on the above platform show favorable linear correlations with those measured using a handheld analytical spectral device (ASD). Their R² values are 0.6052 and 0.6093 and root-mean-square errors (RMSEs) are 0.0487 and 0.1521, respectively. The field-based crop phenotyping platform provides a carrier for high-throughput acquisition of crop phenotypic information.

1. Introduction

Rapid, non-destructive crop phenotyping is key to facilitating the selection and breeding of superior varieties [1,2]. The traditional crop phenotyping methods mainly include artificial sampling and survey, which are inefficient and not timeous, greatly hindering the development of intelligent breeding technologies. With the application of sensors including spectra, images, and laser radars to crop phenotyping [3], the efficiency and accuracy of crop phenotyping have been improved to some extent. However, when using these devices, time-consuming and laborious manual measurement is still needed. These devices are heavy and can only measure at a single point, thus failing to meet the requirements for high-throughput tests of crop phenotypes. In comparison, phenotyping platforms have become an important research direction for high-throughput acquisition of crop phenotypic information because they can carry various types of sensors and collect multi-source data in a short period of time [4].

In recent years, high-throughput field-based crop phenotyping platforms have developed rapidly due to the emergence of automatic and electronic information techniques. According to the working environment, the phenotyping platforms can be divided into indoor and field-based ones. Therein, the indoor crop phenotyping platforms are characterized by high accuracy, strong repeatability, and non-susceptibility to external interference. However, indoor monitoring takes plants out of their growing environment, so cannot monitor some agronomic traits that are shown in the natural population [5,6,7]. By contrast, field-based crop phenotyping platforms have received more attention due to their capability of monitoring relevant crop phenotypic traits on the premise of not changing the growing environment of crops. Field-based crop phenotyping platforms are classified into unmanned aerial vehicles (UAVs), gantry platforms, farm machinery-based platforms, and self-propelled mobile platforms according to their carriers [8]. UAVs have found wide applications in the large-scale and high-throughput crop phenotyping because of their flexibility, wide spatial coverage, and applicability in complex field environments. However, these platforms have a low load-carrying capacity, which limits the number and mass of sensors carried. In addition, the airflow field generated below UAVs when UAVs hover at low altitudes can disturb the canopy, which limits their popularization and application in low-altitude operations [9,10,11,12]. For gantry platforms, a gantry bracket is used to carry sensors for scanning and monitoring along a track fixed on the ground. Such platforms are characterized by strong applicability and smooth operation, while their large investment and high operation and maintenance cost render them inapplicable to large-area mobile operation in fields [13]. Farm machinery-based platforms mainly use large farm machinery including tractors as the driving platform. They are easily realized, whereas, because farm machinery is too large, its field trafficability is poor. Moreover, because most farm machinery is powered by fuel oils, the vehicle and spray lance vibrate violently in the driving process, which is unfavorable to accurate data acquisition [14,15,16]. Therefore, to decrease damage of platforms to crops, reduce cost, and improve the control and measurement accuracy, more researchers have begun to use self-propelled mobile platforms to acquire crop phenotypic parameters [17,18,19,20,21].

Self-propelled platforms can be divided into legged, tracked, and wheeled variants according to their mode of movement. Legged platforms mainly imitate the leg structure of organisms and therefore can mimic the walking of organisms. The BigDog and SpotMini robots developed by Boston Dynamics Inc. in the United States are designed by imitating the moving posture of dogs [22,23]. Legged robots have favorable trafficability while their structures are complex, use complicated control algorithms, and have high development costs. In addition, these robots waggle significantly when moving in fields consisting of soft soils, so they are not suitable for crop phenotyping. Tracked platforms use a tracked chassis as the walking mechanism. Young et al. developed a small tracked phenotyping platform for broomcorn. The tracked chassis was 0.48 m wide, which enabled the platform to move between rows in a broomcorn field [24]. With good traction and adhesion, tracked platforms have high driving stability. However, these platforms fail to realize row-spanning monitoring in a crop field with a small row spacing of plants due to limitations in the width of the crawler wheels. Compared with the two types mentioned thus far, wheeled platforms have advantages including a simple structure, low design cost, high driving stability, and high trafficability in fields, so they can better meet the requirement of crop phenotyping. Xu et al. adopted a small commercial four-wheeled robot (MMP30) and established an automatically navigated phenotyping platform to acquire images of cotton fields [25]. Ingunn Burud et al. used a modular four-wheeled robot (Thorvard II) to carry multi-spectral cameras and RGB sensors to study phenotypic information of wheat and conducted tests on the climbing performance of Thorvard II [26]. Yuan et al. developed a vehicle-mounted phenotyping system for the peanut canopy. In the system, the ground-based lidar and thermal camera and RGB camera were installed on a remotely-controlled four-wheeled vehicle to collect phenotypic data of peanut canopies [19]. Shafiekhani et al. adopted the temperature and humidity sensors mounted on an on-ground self-propelled four-wheeled platform (Vinobot) to monitor phenotypic parameters of individual plants [27]. Tabile et al. developed a four-wheeled self-propelled platform with a large ground clearance of 1.5 m and adjusted the wheel tread to within a range of 1.5 to 2.5 m through lateral movement of wheels by manually pushing the sleeve structure. In addition, they conducted finite element analysis on the obstacle-crossing operation and the performance of the rack on laterally inclined roads to verify the reasonability of the prototype [18,21]. Ruckelshausen et al., developed a four-wheeled field-based mobile platform, for which the ground clearance and wheel tread could be automatically adjusted. They also used the 3-D simulation software Gazebo to conduct dynamic simulation analysis of the obstacle-crossing performance of the robot [28]. However, existing wheeled self-propelled platforms face many problems when operating in fields due to the interaction of environmental factors in fields under high-density planting regimes. A traditional four-wheeled chassis is generally used in wheeled self-propelled platforms; because the axle limits the height of the chassis above ground, the universality of such platforms is greatly limited. The four-wheeled chassis with a large ground clearance to some extent improves trafficability; however, this lifts the barycenter of the platforms, so the stability of platforms decreases significantly, and the platforms vibrate more violently in fields. Moreover, the degree-of-freedom of the platforms increases, and the dead weight grows, which cause lower driving stability, with the addition of devices for adjusting the wheel tread and height above ground. The shadow cast by a closed four-wheeled structure and the vibration of platforms caused by soil heterogeneity also exert certain adverse effects on the crop phenotyping, however, wheeled self-propelled platforms are seldom reported.

To tackle these problems, a rolling adjustment method for the wheel tread was proposed according to the changes in plant row spacing and height of common dry crops. A self-propelled field-based crop phenotyping platform was developed, for which the wheel tread and the monitoring height can be adjusted. It has high trafficability and stability, and therefore provides a reliable field mobile platform for high-throughput crop phenotyping.

2. Materials and Methods

2.1. Design of the Mechanical Structure of the Field-Based Crop Phenotyping Platform

2.1.1. The Structure of the Platform

Numerous crop phenotypic traits can be obtained by analyzing canopy information of crops, so the field-based crop phenotyping platform works by moving spanning rows, which is conducive to high-throughput acquisition of canopy characteristics of crops. A simple open frame is used in the mobile platform, which avoids the adverse influences of the shadow cast by the platform on the monitoring of sensors. The mechanical structure of the platform includes a chassis, a wheel-tread-adjusting device, and a height-adjusting device (Figure 1). To meet the demand for phenotyping of different dry crops in different development stages (corn, wheat, soybeans, etc.), the adjustment ranges of the wheel tread and height of the platform were separately designed to be 2400 to 3200 mm and 1700 to 2100 mm, combining with the row spacing and planting pattern [29]. The platform needs to have favorable trafficability and stability to adapt to the non-structured condition of farmland. The main technical parameters for structural design of the field-based crop phenotyping platform are listed in

Table 1. Technical parameters for structural design of the field-based crop phenotyping platform.

Parameter	Value
Axle distance	2000 mm
Adjustment range of wheel tread	2400 to 3200 mm
Adjustment range of height	1700 to 2100 mm
Total mass	160 kg
Driving speed	0 to 3 m/s
Battery Life	4 h
Obstacle-crossing capacity	≥100 mm
Climbing capacity	≥15°

.

The commonly used chassis of field-based mobile platforms is generally based on three-wheeled and four-wheeled structures, and their advantages and limitations are compared in

Table 2. Comparison of advantages and limitations of the three and four-wheeled chassis.

Characteristics of Chassis	Advantages	Limitations
Four-wheeled structure	High load-carrying capacity	Complex structure
	Strong power	high cost
Three-wheeled structure	Simple structure and low cost	Less appealing appearance
	Easily spans rows High trafficability

. The three-wheeled chassis can adapt to different row spacings and monitoring demands with only a set of wheel-tread- and height-adjusting devices. In addition, the simple and open structure of the three-wheeled chassis can effectively avoid the adverse effects caused by the shadow cast from the closed four-wheeled chassis on the monitoring sensors. Considering these, the three-wheeled chassis is used in the present research.

The front wheel in the walking system of the three-wheeled chassis is a universal wheel that only plays a supporting role, while the two rear wheels are hub-motor driven wheels that move in an individually driven and differential steering mode. According to the relationship between the rigid wheel pressure and deformation (Equation (1)) proposed by Bekker [30], the driving resistance of the wheels is negatively correlated with the width and diameter of the wheels. The row spacing of common dry crops such as wheat is relatively small, from only 150 to 350 mm. Hence, wheels with a width of 100 mm and a diameter of 550 mm are used to reduce the influences of canopy closure in the late development stage of the crop and to lower the barycenter of the platform.

$$F=\frac{{\left(3\frac{F_n}{\sqrt{D}}\right)}^{\frac{2n+2}{2n+1}}}{{\left(3-n\right)}^{\frac{2n+2}{2n+1}}\left(1+n\right){\left(K_c+b{K}_{\phi }\right)}^{\frac{1}{2n+1}}}$$

(1)

where F and Fn separately represent the driving resistance and the vertical load on the wheels; Kc and Kφ separately denote the cohesion modulus and friction modulus; b and D separately refer to the width and diameter of wheels; and n is the settlement index of soils.

The driving system of the three-wheeled chassis directly influences the rideability of the mobile platform, including the obstacle-crossing and climbing performance. Equation (2) indicates that the maximum moment needed by a single driving wheel when climbing a slope of 15° is 128.67 N m. Therefore, the hub motor is a brushless DC motor with rated voltage of 48 V, rated power of 800 W, and maximum torque of 150 N·m.

$$T_1=\frac{\mu mg\cos \alpha +mg\sin \alpha }{3}\times \frac{D}{2}$$

(2)

where T₁ represents the moment needed by a single wheel when the platform climbs a slope; μ refers to the coefficient of rolling friction between wheels and farmland and valued 0.35 here; m denotes the mass of the platform; α is the climbing angle; and g is the gravitational acceleration (9.8 m/s²).

To reduce the vibration and impact caused by uneven farmland, the walking system of the three-wheeled chassis is connected to the vehicle body via vibration dampers. The damper at the front wheel is linked to the rack and those at the rear wheels are connected to the left and right-side arms in the upper parts. The three dampers jointly bear the mass of the upper part of the whole vehicle, so pneumatic damper forks with large load-carrying capacities, light weight, and high sensitivity are used. According to the flatness of farmland and the obstacle-crossing capacity of the platform (not lower than 100 mm), the stroke and load-carrying capacity of the damper forks are 150 mm and 65 to 73 kg. In addition, the prepressing level is 20%, that is, the initial air pressure is 700 kPa. The diameter of the upper tubes and the total length of the damper forks are 28 mm and 650 mm, respectively.

The rear wheels of the three-wheeled chassis are connected to the wheel-tread- and height-adjusting devices via the damper forks, thus forming a gantry structure. The ample space below the gantry ensures the trafficability of the platform in fields. The rack is used to link the gantry and the front wheel, and a battery box is installed on the rack. The rack is only 300 mm wide, which allows the platform to move between rows of crops. The ground clearance is 280 mm, which lowers the gravity center while improving the operational stability of the platform. The schematic diagram for the overall structure of the field-based crop phenotyping platform is illustrated in Figure 2.

2.1.2. Wheel-Tread-Adjusting Device

The function of adjusting the wheel tread to a large extent influences the field adaptability and trafficability of mobile platforms under different planting patterns. Graded mechanical adjustment is generally adopted to adjust the wheel tread of commercially available farm machinery, which needs to use a jack to lift the chassis over ground and adjusts the wheel tread by changing the tightness of the screws. The approach is inefficient and less feasible. Before lifting the chassis over ground, wheels, as the load-carrying parts, encounter large resistance with soils when they move laterally, so the direct forced adjustment faces some difficulty. Therefore, the key to realizing flexible adjustment of the wheel tread is to overcome the excessive resistance encountered during axial movement of the wheels. The wheel-tread-adjusting device on the platform converts the axial sliding of wheels into rolling and changes the sliding resistance to rolling resistance, which dexterously overcomes the overlarge resistance to axial movement of wheels.

The wheel-tread-adjusting device of the platform mainly consists of the left- and right-side arms, a beam, a torque motor, and a horizontal linear actuator (Figure 3 and Figure 4). Therein, the interconnecting component between the output shaft of the torque motor and the damper fork is illustrated in Figure 5. When adjusting the wheel tread, the torque motors on the side arms are used at first to rotate the left and right wheels by 90° on site, so that the rolling direction of the wheels is the same as the direction for adjusting the wheel tread. Then, the horizontal linear actuators on the beam for adjusting the wheel tread are turned on to push (or pull) the sleeves at the left and right ends of the beam and the side arms, to drive the wheels to roll, realizing the adjustment of the wheel tread. After adjustment, the torque motors are turned on to straighten the wheels.

The torque motor is an important component in the wheel-tread-adjusting device. By referring to the Чyдaкoв, Equation (3) [31] and Equation (4), which are widely used in tractor design in China, the DC motor with the maximum torque of 36 N·m is used as the torque motor. The maximum thrust of the linear actuators used is 1200 N, the speed is 5 mm/s, and the stroke is 400 mm.

$$T_2=0.14{\mu }_{S}WR+fWl$$

(3)

where $T_2$ represents the moment needed for wheels to circle around; ${\mu }_S$ is the effective friction coefficient and valued as 0.95 here [30]; W is the vertical load applied by the wheels on the road; f represents the coefficient of rolling resistance and is valued as 0.05 here; l denotes the distance from the center of tire marks to the intersection between the axis of the vertical shaft and the ground and is valued as 0.05 m; and R is the nominal radius of tires, and is 0.28 m.

$$T_3=\mu Wgl\delta$$

(4)

where $T_3$ denotes the torque applied by side arms to the beam (N·m); μ is the coefficient of rolling friction of tires with soils (0.8); $\delta$ is the safety factor of general mechanism (1.5 to 2.0); and l represents the length of the side arms and is set to 1700 mm.

The size of the coupler can be determined according to sizes of the output shaft of the torque motor and the upper tube of the damper fork. The outer diameter, inner diameter, and length of the upper part of the coupler are D₁ = 28 mm, d_l = 14 mm, and l₁ = 80 mm, respectively; those of the lower part are D₂ = 28 mm, d₂ = 14 mm, and l₂ = 100 mm. The coupler is made of ASTM 1045 steel. The beam is produced with a high-strength, high-carbon structural ASTM A283-D steel, and the sleeve in the middle part is hollow square steel measuring 120 mm × 120 mm × 3 mm, while the sleeves at the two ends are also hollow square steel measuring 114 mm × 114 mm × 3 mm. The beam and coupler in the wheel-tread-adjusting device are key interconnecting components of mobile platforms. To ensure that their structural strength meets working requirements, their strength is checked. The allowable stress ${\sigma }_1$ of the beam and the allowable stress ${\sigma }_2$ of the coupler are calculated using Equations (5) and (6) to be 8.91 and 7.33 MPa, respectively. From GB150-1998, the allowable stress $[{\sigma }_1$] of ASTM A283-D steel is 215 MPa, the allowable stress $[{\sigma }_2$] of ASTM 1045 steel is 177.5 MPa, so the beam and coupler both meet the conditions for safe operation.

$${\sigma }_1=\frac{M}{b{h}^2/6}$$

(5)

$${\sigma }_2=\frac{M_{max}}{\frac{\pi \left(D{1}^4-d{1}^4\right)}{16D1}}$$

(6)

where b and h separately represent the cross-sectional width and height of the beam and M_max denotes the output torque of the torque motor.

2.1.3. Height-Adjusting Device

Height adjustment of the platform can not only allow adaptation to changes in the plant height over development stages, but can also meet the adjustment demands of different crop phenotype sensors for the test height. When adjusting the ground clearance of a four-wheeled chassis, two sets of adjusting devices in the front and rear or in the left and right sides of the platform are needed, so the structural stability is poor and vibration amplitude is large. Because the three-wheeled chassis is a T-shaped structure in which the front wheel is on the same vertical plane with the rack, the platform can operate between rows while not damaging the crops. Therefore, one set of height-adjusting devices between the two rear wheels is enough to guarantee the trafficability of the platform in fields, which simplifies the platform structure.

The height-adjusting device of the platform is composed of the vertical sliding rail, the mounting frame of sensors, the small supporting wheel, and the vertical linear actuator, as shown in Figure 6. The vertical sliding rail is connected to the beam via the sleeve, which is welded at the middle part of the beam. The vertical linear actuator is a core component in the height-adjusting device and has a power-off self-locking function at any position during pushing and pulling, and it includes two sets that are installed in the left and right ends of the beam. According to Equation (7), the load borne by a single vertical linear actuator is 490 N, so the maximum thrust of the vertical linear actuator used is 600 N, and its speed and stroke are 5 mm/s and 400 mm, respectively. The ground clearance of the beam is adjusted in the range of 1700 to 2100 mm. When adjusting the height, under action of the small supporting wheel, the positioning dowel on the sliding rail is pulled out to adjust the linear actuators on the side arms and then lift or lower the beam to the needed height. Thereafter, the positioning dowel is fixed on the sliding rail, thus finishing the height adjustment of the platform.

F₁ = m₁gδ/2

(7)

where $F_1$ represents the load borne by a single vertical linear actuator and $m_1$ is the mass borne by the height-adjusting device (50 kg).

2.2. Simulation Analysis

2.2.1. Finite Element Analysis of the Beam and Coupler

Finite element analysis refers to simulating a real physical system (geometric or loading condition) through mathematical approximation. By using simple but interactive factors (elements), it can approximate the real system with infinite unknown quantities using finite unknown quantities. Therein, the static analysis of linear structures is the most common and the most widely used finite element analysis. The main stressed component on the crop phenotyping platform is the gantry formed by the beam and the left- and right-side arms, in which the beam is an important load-carrying component. The coupler is adopted to transmit torque between the torque motor and the wheels, and is an important component in the wheel-tread-adjusting device. To ensure the safe operations of the mobile platform, the linear static analysis needs to be conducted on the coupler and the beam.

The 3-D models of the coupler and beam are imported into the ANSYS Workbench software, which is followed by mesh generation, applying constraints and loads, and solving the models. The middle part of the beam is fixed while moment of 932.96 N·m along the forward and backward directions of the beam is applied to sleeves at the two ends. The solution results are shown in Figure 7; the maximum stress on the beam is at the bending positions at the two ends of the sleeve. The maximum stress is 61 MPa ≤ $\left[{\sigma }_1\right]$ MPa, so the strength of the beam meets design requirements.

The shaft shoulder in the middle of the coupler is fixed. Because the maximum torque from the torque motor connected to the coupler is 36 N·m, torque of 36 N·m is applied to the coupler and the two sides of the key slot are also stressed (Figure 8). The stress on the coupler is concentrated at the key slot, and the maximum stress is 65.243 MPa ≤ $\left[{\sigma }_2\right]$ MPa, so the coupler also satisfies the working requirement.

2.2.2. Dynamic Simulation Analysis

The obstacle-crossing performance and climbing performance are key factors that affect the movement stability and trafficability of the platform. The field-based crop phenotyping platform is large and needs to work on rough farmland tracks. Therefore, to avoid slipping, it is necessary to conduct dynamic simulation analysis on the obstacle-crossing capacity and maximum climbing angle of the platform, so as to better guide the practical application.

(1)

Simulation analysis of the obstacle-crossing capacity

The 3-D entity model of the platform and the road surface model are established in Pro/E and then imported into ADAMS, in which the material attributes of parts are set. The road surface is set as soft soils, and the constraints, friction, motion, and load are added to various parts. According to requirements of the design parameters, the width and height of the obstacle on the road are both set to 100 mm. The vibration amplitude and vibration acceleration are important indices for measuring the obstacle-crossing performance, so the vibration acceleration and vibration amplitude vertical to the ground in the middle part of the beam are measured. Figure 9 displays the obstacle-crossing simulation.

The post-simulation processing is shown in Figure 10. When the crop phenotyping platform passes an obstacle, the vibration acceleration in the middle part of the beam along the height direction of the platform is basically not larger than 0.5 m/s², except for at some points where an abrupt change may occur. The maximum vibration amplitude is 62.5 mm, which is 3.68% of the height of the platform. In addition, the platform is self-restored to smooth running immediately after surmounting the obstacle.

(2)

Simulation analysis of the climbing angle

It is supposed that the platform climbs a slope at a uniform speed and the air drag, inertial resistance, and rolling resistance are ignored. Figure 11 is the schematic diagram for forces on the crop phenotyping platform when climbing a slope.

According to the static analysis during climbing, the limit equilibrium equation under conditions that the driving wheels do not undergo longitudinal slip is

Gsinφ − Fφ = 0

(8)

where G, F, and $\mathsf{\phi }$ represent the platform mass, driving force of driving wheels, and the maximum climbing angle, respectively.

The model of the whole platform and the mobile slope are assembled in ADAMS, as shown in Figure 12. The slope is formed from soft soils and contact forces are separately applied to the three wheels of the platform and the slope. The dynamic and static friction coefficients are separately set to be 0.035 and 0.8. According to the setting, the platform moves on the slope at a speed of 0.8 m/s, during which one end of the slope is fixed while the other end rotates around the concurrent axes. The slope angle is increased at a constant rate of 1° s⁻¹ and, theoretically, the angle at which the platform initially slips on the slope is the maximum climbing angle.

Figure 13 shows the relationship between the slope angle and the Y-coordinate in the vertical direction of centroid of the platform. The coordinate of the centroid of the platform along the height direction is the ordinate, while the slope angle is marked on the abscissa. As shown in the figure, when the slope angle is about 31°, the ordinate changes abruptly, that is, the platform slips at the slope angle. To verify the climbing angle, slopes with angles slightly larger or smaller than 31° are set and the simulation results are displayed in Figure 14, in which the abscissa and ordinate separately represent the simulation time and coordinate along the height direction of the platform. On the slope with an angle slightly smaller than 31°, the coordinate changes smoothly, with the curve showing a certain slope, that is, the platform can climb over the slope without slipping. In the simulation of the slope with an angle slightly larger than 31°, the platform slips at 18 s. Therefore, the maximum climbing angle of the crop phenotyping platform is about 31°.

2.3. Tests of the Field-Based Crop Phenotyping Platform

2.3.1. Test Design

• Performance tests of the field-based crop phenotyping platform

Performance tests of the field-based crop phenotyping platform include three groups of tests, separately on the driving smoothness, obstacle-crossing performance, and climbing performance (

Table 3. Design of performance tests of the field-based crop phenotyping platform.

Test	Test Design	Test Index
Smoothness test	Specifications of the platform (wheel tread × height) are 2.4 m × 1.7 m, 2.4 m × 2.1 m, 3.2 m × 1.7 m, and 3.2 m × 2.1 m.	Vibration acceleration, vibration amplitude
Test on obstacle-crossing performance	The height and width of the obstacle are both 100 mm.	Vibration amplitude
Test on climbing performance	The slope angle is about 15°.	Whether it can be parked or not

). The vibration acceleration and vibration amplitude to some extent reflect the stability and operation smoothness of the platform, so they are used as measurement indexes. In the smoothness test of the platform, the control variable method is used to measure the vibration accelerations of the platform with four different specifications to compare influences of the specifications on the smoothness of the platform. The four specifications include 2.4 m × 1.7 m, 2.4 m × 2.1 m, 3.2 m × 1.7 m, and 3.2 m × 2.1 m (wheel tread × height of beam). In the test on obstacle-crossing performance, the obstacle is set to have a height of 100 mm and a width of 100 mm to measure the vibration amplitude of the platform when crossing the obstacle. Limited by the test site, the angle of the slope used in the test on climbing performance is about 15° to verify the parking performance of the platform during climbing.

2.

Field tests

Field tests of the platform mainly include two parts: field vibration tests and wheat-growth monitoring tests. In the field vibration tests, the vibration accelerations at two measuring points on the platform during straight driving at a constant speed in a field were measured to determine the optimal loci for installation of sensors. The wheat-growth monitoring tests were conducted in a test base in Baipu Town, Rugao City, Jiangsu Province, China, and the wheat variety Shengxuan 6 was planted with a row spacing of 200 mm. Nitrogen fertilizer at a dosage of 360 kg/ha, phosphate fertilizer at 135 kg/ha, and potassium fertilizer at 220 kg/ha were applied together with the base fertilizer for a single treatment. Other cultivation management measures were same as those in general high-yield fields. The installation locations of sensors for measuring phenotypes were determined by results of field vibration tests. Sensors were mounted on the platform, which then moved in a straight line at a fixed speed to measure growth parameters of wheat.

2.3.2. Test Devices

(1)

Field-based crop phenotyping platform

The tests were conducted using the crop growth sensor based on a passive light source and a DH5902N dynamic signal test and analysis system mounted on the field-based crop phenotyping platform. The field-based crop phenotyping platform is illustrated in Figure 15a.

(2)

Crop growth sensor

The crop growth sensor uses sunlight as the light source and a dual-band detection lens with an angle of view of 27° in the tests. The sensor can provide the real-time output of the NDVI and RVI of crop canopies. By using a photosensor to realize high-fidelity energy conversion from optical energy to electric energy, the sensor structure can be divided into an uplink and a downlink optical sensor. The uplink optical sensor is used to get radiation information of sunlight at wavelengths of 730 and 815 nm, while the downlink one is applied to receive radiation information of reflected light from crop canopies at the corresponding wavelengths. The tests need to be conducted in calm and cloudless fine weather and the measurement height should be fixed to be 1.0 to 1.5 m above the crop canopies. The field of view is aimed vertically downward, and the crop canopies remain relatively at rest. It is convenient to perform secondary development for the crop growth sensor and the test results of the canopy reflectance of wheat are well correlated with those of the FieldSpec ASD [32]. The crop growth sensor is depicted in Figure 15b.

(3)

ASD FieldSpec HandHeld 2

The ASD FieldSpec HandHeld 2 is developed by the Analytical Spectral Devices in the United States and can be used to acquire reflectance spectra of different objects including crops, marine organisms, and minerals. The spectrometer data are directly stored in the device and the results are accurate. However, secondary development is challenging for the device. Its measurement wavelength is in a range of 325 to 1075 nm, and the wavelength accuracy, spectral resolution, and angle of view are ±1 nm, less than 3 nm, and 25°, respectively (Figure 15c).

(4)

DH5902N dynamic signal test and analysis system

The DH5902N dynamic signal test and analysis system is composed of a DH5902N rugged data acquisition instrument produced by Jiangsu Donghua Testing Technology Limited Liability Company, Jingjiang, China, a 1A314E three-way piezoelectric acceleration transducer, and a portable computer. The axial sensitivities of the 1A314E three-way piezoelectric acceleration transducer in X, Y, and Z-axes are 100.20, 98.71, and 95.14 mV/g, respectively. The measurement range, frequency range, and mass of the sensor are ±50 g, 1 to 6000 Hz, and 15 g, respectively. The DH5902N dynamic signal test and analysis system is mainly applicable to stress and strain tests and can accurately measure multiple physical quantities including force, pressure, and displacement by coordination with various bridge sensors. After being fixed at measurement points, the acceleration transducers output signals via the signal lines to the data acquisition channel of the data acquisition instrument. Based on the WIFI function, the computer is wirelessly connected with the data acquisition instrument, as shown in Figure 15d.

2.3.3. Test Methods

The junction of the beam and the side arm and the middle part of the beam are two key connection points of the platform. Because the mounting frame of sensors is on the upper end face of the beam, the vibration acceleration and vibration amplitude at the two connection points can more truly reflect the influences of the platform on the sensors for measuring phenotypes mounted thereon. Therefore, the above connection points were set as measuring points. In the smoothness test, the acceleration transducers were fixed at pre-set measuring points (the middle part of the beam and the junction of the beam and the side arm) with 502 glue. Platforms of different specifications were set to move at a constant speed of 0.5 m/s on flat ground and the data acquisition instrument was used to attain vibration accelerations of the platform along the direction vertical to the ground and the heading direction of the platform. The sampling frequency was set to 1000 Hz.

In the test on obstacle-crossing performance, the acceleration transducer was glued in the middle part of the beam and the platform crossed the obstacle at a constant speed of 0.5 m/s. In the process, the vertical acceleration was measured at a sampling frequency of 1000 Hz.

In the test on climbing performance, the platform was driven to climb a slope and park on the slope after driving on a flat road at a constant speed of 0.5 m/s to verify the climbing performance of the platform.

The positions of acceleration transducers in the field vibration tests are shown in Figure 15a. The platform was driven in a straight line at a constant speed of 0.5 m/s. In the process, the vibration accelerations at the measuring points vertical to the ground and along the heading direction of the platform were obtained (the sampling frequency was 1000 Hz).

The wheat-growth monitoring tests were conducted to check whether the designed field-based crop phenotyping platform equipped with crop growth sensors could efficiently, non-destructively, and accurately obtain the wheat growth information. The tests were separately performed in the tillering stage, early jointing stage, and late jointing stage of wheat, in calm and cloudless fine weather, from 10:00 a.m. to 14:00 p.m. To avoid influences of the edge effect of large fields, the platform was set to move in a straight line at a constant speed of 0.5 m/s along the planting direction in the field and the middle of the field was randomly measured. Synchronously, the ADS FieldSpec HandHeld 2 was used to acquire the NDVI and RVI of the wheat canopy.

2.3.4. Data Analysis

Vibration acceleration data obtained in the tests were processed and analyzed using the signal processing and analysis module in the DHDAS (Donghua test real time data measurement analysis software) dynamic signal acquisition and analysis system. The phenotype data measured in the field tests were subjected to statistical analysis using the EXCEL2016 software and the correlations in the data were evaluated using the determination coefficient and the root-mean-square error (RMSE).

3. Results

3.1. Smoothness

The results of the smoothness tests of the platform during driving are shown in Figure 16 and Figure 17. With the increases in the wheel tread and height of the platform, the accelerations in the middle part of the beam and the junction of the beam and the side arm did not change to any significant extent. The accelerations at the two measuring points along the heading direction were greater than those in the vertical direction and none exceeded 0.50 m/s². Except for at some points, the vibration accelerations at the two measuring points in the vertical direction were less than 0.50 m/s² when the wheel tread and height were adjusted to their maximum values. Moreover, the accelerations along the forward and backward directions were always within 0.50 m/s². Therefore, the crop phenotyping platform was slightly influenced by the wheel tread and beam height when driving on flat ground, indicative of favorable driving smoothness.

3.2. Obstacle-Crossing Performance

Test results of the obstacle-crossing performance are illustrated in Figure 18. Except for the points at which an abrupt change occurred, the vertical accelerations at the measuring point in the middle part of the beam were all below 0.5 m/s² and the maximum vibration amplitude was 88.7 mm in the obstacle-crossing process. The platform self-restored to smooth movement 8 s after crossing the obstacle, indicating that the platform has certain obstacle-crossing and damping capacity.

3.3. Climbing Performance

Test results of the climbing performance suggest that the platform can climb over and park on slopes while it does not roll over or slip. Therefore, the platform has the capacity to climb a slope with the angle of 15°. Figure 19a,b show the climbing test site and the parking of the platform during climbing.

3.4. Field Tests

Field vibration test results are shown in Figure 20. The junction of the beam and the side arm vibrated more intensely than the middle part of the beam, and the vertical vibration amplitudes at the two measuring points were greater than those on flat ground. This is because many farmland roads are rough and uneven due to the heterogeneity of soils, and therefore the vertical vibration amplitude increases significantly. At 13 s, when the platform climbed over an obstacle in the field, the vertical vibration amplitude at the junction of the beam and the side arm reached the maximum of 248.3 mm, which is 14.6% of the height of the platform. The maximum vibration amplitude in the heading direction of the platform was 102.1 mm, which is 4.9% of the length of the platform. After 10 s, the vibration amplitude remained within 30 mm, which is only 1.8% of the height of the platform and the small vibration amplitude is attributed to the non-structured, soft nature of such farmland tracks. When crossing the same obstacle, the vibration amplitude in the middle part of the beam was smaller than that in the junction of the beam and the side arm and the vibration amplitude along the heading direction of the platform was smaller than 30 mm. The vibration caused by uneven farmland tracks is transmitted to the junction of the beam and the side arm mainly via the two side arms and then sent to the middle part of the beam; energy is dissipated in the transmission process, so sensors should be mounted at positions as near as possible to the middle part of the beam.

Figure 21 shows the linear fitting results for the NDVI and RVI, measured separately using the crop growth sensor mounted on the platform and the ASD FieldSpec HandHeld 2 spectrometer in the tillering stage, early jointing stage, and late jointing stage of wheat in the field.

4. Discussion

Field-based crop phenotyping platforms are characterized by a high upper limit of load and long endurance; therefore, they have received more attention in the current research and application of precision agriculture. However, some existing platforms have poor trafficability in the field and do not consider cultivation and agronomic measures, so they find it difficult to adapt to changes in the plant height and row spacing over development stages of crops [16,26,33,34]. Therefore, the field-based crop phenotyping platform with automatic adjustable wheel tread and height was designed according to cultivation and agronomic requirements of wheat by taking dry crop wheat as the main research object. In the meantime, a series of tests were conducted to verify that the main structural design of the platform can meet the demand for field-based crop phenotyping and the platform has strong trafficability and high stability in fields.

The vibration amplitude of the platform when crossing an obstacle simulated based on ADAMS was 62.5 mm, while the maximum vibration amplitude in the test on obstacle-crossing performance was 88.7 mm. The difference is because components are rigidly connected in the simulation model and these components do not undergo relative displacement—whereas mounting clearance is present between components of the real platform, which may cause superposition of vibration. The road model was relatively ideal in the dynamic simulation using ADAMS, so effective road spectra in the field will be established in the future to render the simulation a better fit to actual conditions. Compared with the vibrations of tractors in the field, the vibration accelerations at measuring points in the field test were significantly reduced [35], which is more conducive to the stability and accurate monitoring of sensors.

5. Conclusions

(1)

According to differences in the row spacing and plant height of common dry crops, a rolling adjustment method for the wheel tread was proposed and a self-propelled three-wheeled field-based crop phenotyping platform was developed. The platform has functions allowing for the adjustment of the wheel tread and the height of the monitoring platform. It has strong trafficability in fields, high stability, and a wide application scope, and therefore provides a reliable carrier for high-throughput acquisition of crop phenotypic information.

(2)

ANSYS finite element analysis results reveal that the strength of the beam and coupler meets safe working conditions. ADAMS dynamic simulation results indicate that when the platform crosses an obstacle of 100 mm in height, the accelerations are less than 0.5 m/s²—except for at points at which abrupt change occurs—and the vibration amplitude is less than 4% of the height of the platform. The maximum climbing angle is about 31°. The results indicate that the virtual platform has favorable obstacle-crossing and climbing performance and high driving smoothness.

(3)

Tests indicate that the platform can ride smoothly over flat ground and that different wheel treads and heights of the beam exert only slight influences on the vibration amplitude. Except for several points at which an abrupt change occurs, accelerations at other measuring points are less than 0.5 m/s². When climbing over an obstacle with a height of 100 mm, the vertical vibration amplitude in the middle part of the beam is 88.7 mm. The climbing capacity is ≥15°. When driving in a field, the vibration amplitude in the middle part of the beam is lower than that in the junction of the beam and the side arm, so sensors need to be installed near the middle part of the beam. Tests in fields of wheat suggest that the NDVI and RVI measured using the crop growth sensor have close linear correlations with the data measured using the handheld ASD, with R² of 0.6052 and 0.6093 and RMSEs of 0.0487 and 0.1521, respectively.

6. Patents

China patent: Field crop phenotypic monitoring robot. Patent number: CN 109466655B [36].