A Prototype Crop Management Platform for Low-Tunnel-Covered Strawberries Using Overhead Power Cables

Abstract

The continuous and reliable operation of autonomous systems is important for farm management decision making, whether such systems perform crop monitoring using imaging systems or crop handling in pruning and harvesting applications using robotic manipulators. Autonomous systems, including robotic ground vehicles, drones, and tractors, are major research efforts of precision crop management. However, these systems may be less effective or require specific customizations for planting systems in low tunnels, high tunnels, or other environmentally controlled enclosures. In this work, a compact and lightweight crop management platform is developed that uses overhead power cables for continuous operation over row crops, requiring less human intervention and independent of the ground terrain conditions. The platform does not carry batteries onboard for its operation, but rather pulls power from overhead cables, which it also uses to navigate over crop rows. It is developed to be modular, with the top section consisting of mobility and power delivery and the bottom section addressing a custom task, such as incorporating additional sensors for crop monitoring or manipulators for crop handling. This prototype illustrates the infrastructure, locomotive mechanism, and sample usage of the system (crop imaging) in the application of low-tunnel-covered strawberries; however, there is potential for other row crop systems with regularly spaced support structures to adopt this platform as well.

1. Introduction

Automation in agriculture is becoming increasingly necessary due to several recent challenges, such as the growing global population, yet shrinking agricultural labor force, the need for precision management to reduce farm operation costs, and rising customer expectations for higher-quality staples. Numerous methods have been developed to automate crop monitoring and management with significant progress in the use of drones and unmanned ground vehicles. Drones can cover large areas quickly, providing a high-level view of crop status, while ground vehicles can provide more detailed insights into individual plants and perform specific management tasks. Automation systems are also being developed for controlled environments, such as indoor farming and high tunnel systems, where the system can be ground-based [1] or mounted on overhead rails [2,3]. While indoor farming has a higher capital start-up cost, the controlled environment can facilitate the extensive use of sensors, automation, and robotics to optimize yield and minimize labor [4]. Another approach has been to use wire-borne platforms, such as the Tarzan robot, which uses a brachiating movement on ropes for outdoor agriculture monitoring [5]. One detail that is often overlooked is how to ensure the continuous operation of crop monitoring systems. Batteries are the preferred option, with a widely used commercial ground platform—the farm-ng Amiga—claiming to offer a runtime of three to eight hours, depending on usage [6]. Other heavier unmanned machinery may require a diesel-powered engine, such as the AgXeed AgBot series, or battery charging gas generators, such as the Agrobot E-Series [7,8].

The crop management technologies discussed above offer distinct advantages, but also face limitations that restrict their use in diverse agricultural environments. Drones are limited by the flight time and struggle to provide detailed, ground-level data on individual plants. Smaller agricultural ground vehicles, while capable of precise navigation, can be hindered by challenging terrains. Heavy agricultural machinery, though effective for power-intensive tasks and capable of overcoming complex terrains, need a high initial capital investment and can contribute to soil compaction. This work proposes a solution that attempts to bridge these gaps by enhancing the cable-borne robotic system approach with wheeled locomotion over electrified cables. Similar platforms have been investigated for powerline inspection applications, such as the LineRanger [9], Meta’s Bombyx for wrapping fiber-optic around powerlines [10], and the Mini LineFly by the Preformed Line Products (PLP) Company for installing bird diverters on powerlines [11]. These platforms are built for existing overhead powerlines, but none can be found for agriculture applications, let alone on how one might install the infrastructure and cabling for such applications. Many of these platforms use batteries for operation, although some commercial units can last many hours depending on the task [12]. Others, like Bombyx, use gas power motors or can be towed with a rope by a person on the ground. Many of these platforms can also be heavy, for example, LineRover weighs near 23 kg [12]. For handling small fruit crops or imaging them, such heavy machinery may not be necessary, and a lighter platform can mean reduced costs per unit and an increase in locomotive speed to cover more acres quickly.

In this work, a robotic platform is proposed which uses overhead cables for both navigation over crops and power delivery to the system. The compact 3.25 kg system is adaptable for both indoor and outdoor agriculture environments that use regularly spaced crop support structures. The current implementation allows the platform to reach a speed of 4.5 km/h. Its modular design offers flexibility for swapping components to complete different agricultural tasks, such as monitoring and plant manipulation. The use of cables for power delivery is intended to replace the need for batteries or other heavy power supply units, while also providing a reliable navigation mechanism over crops. The overhead navigation also helps to reduce soil compaction, and the robot’s operation is not affected by the ground terrain conditions.

The goal of this work was to develop a proof-of-concept cable-driven robot. To achieve this, the objectives included the design and construction of the cable-driven robot and modifications to row crop infrastructure for supporting the cables. A method to draw power from the cables was implemented, and a common crop management task was tested with the system—in this case, crop monitoring using a machine vision system, but the platform has the capacity for other management tools to be mounted as well, such as robotic manipulators. The concept was then demonstrated for a low-tunnel strawberry bed system. In this planting system, strawberry beds are covered by low, semi-permanent plastic tunnels supported by metal hoops spaced approximately five feet apart (Figure 1). The plastic and metal hoops covering the beds make it difficult for either drones or ground vehicles to effectively monitor the plants or navigate over them. Robotic ground vehicles often use the spacing between the beds for their wheels to traverse across the rows with the center of the vehicle passing right above the plants. In Figure 1, the height of the metal hoops would likely interfere with the bottom of the vehicle, and even if the vehicle was raised higher, the plastic would still prevent the top-down visibility of the plants; therefore, one solution is to operate the vehicle within the tunnels. In this work, the metal hoops provide the necessary support structure for the cables with a suspended robotic platform to traverse inside the tunnels.

For the remainder of this paper, the sections are as follows: In Section 2, a high-level overview of the methods used to develop the system and testing is given. More detailed instructions on how the platform was built, including CAD files, a bill of materials (BOM), and code are provided in the Supplementary Materials [13]. In Section 3, the results on the performance of the prototype and the cable support infrastructure are evaluated, and in Section 4, the observed challenges and shortcomings of the system are discussed, with thoughts on alternative methods that could be used to overcome these challenges.

2. Materials and Methods

There were two parts to developing the system: the infrastructure/cabling setup and the robotic platform itself. The cost of the robotic platform with the imaging system was approximately USD 1071. The cost of the support structures for one row was USD 170 + USD 1.28 per meter of powerline cabling. A detailed view of the robot structure with interconnected components and assembly instructions can be found in the Supplementary Materials.

2.1. Infrastructure and Cabling Setup

Structural support is used in row crops for numerous reasons and often includes wooden end-posts anchored into the ground with evenly spaced posts in between. They help support trellis wires running along the rows, as shown in Figure 2A for an orchard and vineyard. For the strawberry row, a similar structure was implemented. Two wooden posts were anchored into the ground at an angle at opposite ends of the row, with one end having winches attached to the post to tension the power cables (Figure 2B) and the other end having two eye-hooks drilled into the wood to hold the other end of the cables (Figure 2D). For the in-row support structure, custom 3D-printed parts were attached to the steel hoops that typically support plastic tunnels. The cables then rested on these supports (Figure 2C). The full setup is shown in Figure 2E during one of the data collection trials with the imaging system. A DC bench power supply was used to electrify the cables which powered the robot platform, with positive voltage and ground annotated on this image.

The cables chosen were aluminum conductor steel-reinforced (ACSR), commonly used for overhead powerlines. They consist of a steel wire core, which provides tensile strength, and aluminum wire strands wrapped around the core for conductivity. A range of wire diameters is available for consumer purchase, with 0.257” (Code Name: Swanate) selected for this application due to its ability to support high ampacity for power-intensive uses, while remaining light enough for easy tensioning. The ACSR cables can be insulated or bare; the bare option was chosen with the assumption that metal-to-metal contact between the robot wheel and cable would provide an effective way of delivering power to the robotic system. A few obstacles to this assumption became apparent later on in system development, which are stated in the discussion section of this work, along with potential solutions. The supporting structures for the cables attached to the steel hoops and the robotic platform body were 3D-printed using PETG filament, chosen for its good balance of heat resistance, ease of printing, and cost-effectiveness compared to other filaments. The experiments took place at the Cornell Agricultural Experiment Station in Ithaca, NY, USA. The cabling and the structures were left in the field for six months between July and December 2024, while testing and experiments took place. During this period, temperatures ranged from −5 °C to 33 °C with a variety of weather conditions from hot, humid, and heavy rainfall to dry, windy, and freezing temperatures. No damage or deformation of the structural support materials was noticed during this period. Tests were conducted to collect navigation, power supply stability, and image data of the plants between July and October, with weekly sessions lasting up to four hours. During November and December, the condition of the support structures was mainly checked for durability and any damage from the freezing temperatures. The robotic platform itself was not tested in rain or snow conditions; however, it is assumed that safe operation is possible during these conditions if the plastic is brought down to the edge of the beds, providing both insulation and barrier from the outside weather conditions.

2.2. Robot Platform

The robotic platform moving along the cable has three main components: power delivery, drive system, and targeted application use (Figure 3A). Power delivery in this design assumes direct current (DC) at 16 volts (V) running through the cables. Other locomotive transportation systems use alternating current (AC) or DC, with AC offering the advantage of easily stepping up the voltage for long-distance power transmission, while DC requires fewer components to convert power from the source to the electrical components that utilize it. As the aluminum wheels of the robot roll along the cable, the current is transferred from the wheel’s surface to the axles and to the rest of the system (Figure 3B). The wheels, however, make intermittent connections with the axle. This creates noise in the power supply being supplied to the robot and needs to be filtered. A decoupling capacitor and a low-pass filter first stabilize the DC voltage, which is then transferred to several DC step-down converters for additional filtering, and finally as an output to various voltage requirements by the system. To test the stability of power being drawn from the cables, an electronic load tester [14] was attached to the system to draw power, and a digital oscilloscope [15] continuously recorded the voltage of the system at 375 kilosamples per second, while the platform was in motion on the cables.

For the drive system, two stepper motors move the robot along the power cables. The drive wheels were 3D-printed with TPU filament material. The surface of the wheels has a u-channel shape slightly smaller than the diameter of the cable to provide better grip on the cable. This helps to reduce wheel slippage, especially when traveling at an incline. Each drive wheel is equipped with two aluminum wheels on either side that rotate along with the drive wheel, providing additional stability and transferring power from the cables to the robot. This power transfer mechanism is similar to electric trains that use pantographs or contact shoes to collect electrical current; however, this approach uses wheels for current collection since the rolling motion on the cables creates less friction (which the motors then do not have to overcome) and less wear and tear on the cables. Extension springs are used to keep the aluminum wheels in contact with the cables.

An AS5600 magnetic encoder sensor board was mounted on the back of one of the stepper motors. By continuously reading the motor shaft position, the speed was then determined by the amount of time it took for the wheel to make one rotation. Instructions on how to mount the encoder sensor, including the 3D print files for the motor bracket, from [16] were followed. This setup can be used for recording wheel odometry to spatially align remote sensing data that is captured. In one of the trials, an LiDAR sensor (model RPLIDAR A1, Slamtec, Pudong New District, Shanghai, China) was mounted on the platform to register point cloud data of the strawberry bed and plants.

For the targeted application, an imaging system was developed for the top-down viewing of the plants. The camera was an embedded vision MIPI camera from e-con Systems (Model e-CAM24_CUNX_H01R3, Chennai, Tamil Nadu, India) connected to a Jetson Nano computer module (Model B01 Developer Kit 4GB, NVIDIA Corporation, Santa Clara, CA, USA). The camera was chosen for its compact form factor, global shutter with an external trigger feature, cost-effectiveness compared to other industrial brands, and the CSI-2 interface, which has a fast and reliable image transfer protocol with embedded computer modules like the Jetson Nano. An LED flash system was built following designs from [17,18], which synchronized to the camera’s shutter using the external trigger (Figure 4). The maximum speed of the system driving on the cables was tested to reach 4.5 km/h, which can cause motion blur in the images; therefore, the flash system allowed us to shorten the shutter duration, while keeping the image foreground lit for obtaining a better image quality. To test the system, one section of the strawberry row was imaged with the platform accelerating to 4.5 km/h with a camera framerate of 10 fps.

3. Results

3.1. Power Delivery Using Cables

The analysis of power transfer from the cables to the robot platform is shown in Figure 5. Figure 5A shows the voltage supply when there is no signal filtering as the platform is in motion on the cables; this level of noise would not have allowed for any electrical components to function properly. Figure 5B shows the voltage levels after filtering, with the three output voltage signals drawing a continuous 4 amp (A) load, while the platform is in motion. The 16 V plot is the raw signal after high- and low-frequency filtering. The 5 V and 12 V lines are the result of the filtered signal going through the additional stepdown DC regulators. It is assumed that the filtered 16 V will go through an additional filtering stage for the targeted application. In this work, the targeted application is the imaging system, and the additional filtering is a DC-DC boost converter used for the flash unit that raises voltage from 16 V to 50 V.

3.2. Locomotion and Speed Testing

Video S1 in the Supplementary Materials shows the platform accelerating to the max speed of 4.5 km/h before decelerating [13]. The results of the speed test are shown in Figure 6A, with a time series graph of wheel velocity. Video S2 in the Supplemental Materials shows the platform during an imaging trial of the strawberry plants as it rolls over the cables and the steel hoop support structures at a slower speed [13]. Figure 6B demonstrates how wheel odometry using a magnetic encoder can be combined with other sensors, such as LiDAR, to register data points. The image shows one of the trials with the LiDAR sensor mounted on the bottom of the platform traversing over the strawberry bed and mapping the structure of the bed and the plant canopy. A QR code was placed at the beginning of the row, and such a tag could be used for each subsequent row for the global registration of the sensor data. Using wheel odometry as the sole localization sensor to register the sensory data is often less reliable for ground vehicles as the wheels can slip in the X and Y directions and during turn adjustments to stay on a straight path. With the cable drive system, the robot is more restricted in its movement, traveling only forward and back; however, some errors can still be introduced due to the sagging of the cables, which can create a vertical sinusoidal travel path rather than a straight one.

3.3. Sensor Data Gathering and Processing

Figure 7 shows the results of the imaging platform while traversing over a section of the strawberry row at 4.5 km/h. As shown in Figure 7A, the auto-exposure feature of the camera was used to capture this image, and Figure 7B presents an image with manual exposure with synchronized flashes. In high-speed imaging, the flashes do help with lessening the motion blur effect; however, the effect was not noticed at lower speeds with the auto-exposure setting. For ground vehicles, rough or uneven terrains can still cause the blur effects in images, but given the smooth motion over the cables, a flash system may not be necessary for the cable system unless it is needed to account for lighting variability in the images. Another benefit of the flashes in this experiment was that regardless of the exposure setting of the camera or time of day, as seen in Figure 7C,D, external lighting can help to penetrate deeper into the canopy (Figure 7E). Combined with the shorter shutter duration, the image with active lighting also appears to reveal more granular texture details of the plant canopy. There are also opportunities to use this active lighting method for multispectral imaging by adding LEDs of different wavelengths.

4. Discussion

This prototype was tested intermittently during a six-month period in July–December. The main advantages of this system over a ground vehicle would be the increased navigation speed to cover rows quickly, its compact design to traverse inside low-tunnel structures, and the reduction in soil compaction. Its main advantages over drones are a higher payload (though less than a ground vehicle), a longer operation time, and its capability to offer both the high-speed ground coverage of crops, and at the same time, perform crop-level manipulation tasks.

Given these advantages, some limitations were also realized during the development and testing of the system that would require either regular maintenance or future improvement. These limitations are given below, including potential solutions.

Several advantages were given regarding the system compared to drones and ground vehicles. Some of the disadvantages compared to those systems are that while the cable-driven vehicle is lightweight and makes no ground contact, it has a limited payload capacity since adding more weight will cause the supporting cables to sag. For a low-tunnel system, this may not be as much of a concern due to the steel hoops being typically spaced every 1.5 m. Additionally, the proper arrangement of components on the robot is important to ensure the center of gravity is aligned; otherwise, the unit may start to tilt. One solution which should be implemented in the future is to move the two power cables as far apart as possible from each other to provide better stability for the vehicle. To achieve this, in reference to Figure 2, the eye hooks can be placed on the side of the wooden posts, and a new 3D print design should be made so that the u-channels holding the cables can be attached to the sides of the steel hoops. The drive panels should then be flipped so the wheels face outward, with additional metal rods attached near the top of the drive panels to firmly hold the two pieces together.

The cables and the supporting structures can be re-used for multiple seasons; however, the initial setup and end-of-season teardown requires additional labor. Crops with permanent support structures, such as orchards or vineyards, which the cables can remain on indefinitely may be more suited for such a system, unless a method to tension the cables using readily available structures for the low-tunnel system, such as the metal pipes at the ends of the strawberry bed rows (Figure 1), can be used. The additional wooden posts at the end of the rows could then be removed.

One unexpected result during testing was that because of intermittent connection between the aluminum wheels and the axles; during high-current draws, there seemed to be electrical arcing in the small gap between the axle and the wheel bore. Over time, this leads to a layer of black residue (likely oxidization) that forms over the surface of the aluminum axle and the bore. This layer will need to be wiped down if it becomes thick enough to block the electrical connection. An acidic solution and a mild abrasive material can be used to scrub the residue off. This issue was an unexpected setback to making the system more autonomous and resolving it will be crucial for future development. Although DC voltage connectivity between the power source and the platform is ideal for efficient power delivery with minimal electronics, one solution would be to run AC voltage at a much higher frequency (~kHz range) through the cables and allow for the platform to draw power using electromagnetic induction (EMI) as it navigates over the cables. An additional advantage to this approach is that the cables can be shielded, since EMI power transfer can be contactless. This would also alleviate concerns regarding bare live wires running through a crop field.

Future work would include implementing a turning feature for the vehicle to navigate between rows. Rather than using RTK GPS systems, there is an opportunity to use only wheel encoders, vision system, and QR tags placed at beginning of the rows for global sensor data registration. Additionally, an improvement could be made for the tensioning mechanism of the cables. Such mechanisms and tools exist for trellis wires, but those are much thinner than the cables used in this system. Finally, the improvement of the power delivery method will need to be conducted for continuous operation, whether through the development of innovative wheel-to-wire contact without electrical arching or with the use of an EMI and AC power supply.

5. Conclusions

In this work, a prototype vehicle for indoor and outdoor row crop management was presented, in which the robotic platform uses overhead power cables for operation and locomotion, an alternative to ground vehicles or drones, which both require batteries to operate. The design and construction of the concept vehicle, including support structures for cabling, were tested on a strawberry bed low-tunnel system. Low-tunnel systems have an enclosure which makes it challenging for drones or other ground vehicles to monitor crops. The platform was tested for power delivery, locomotion, and data collection, while traversing over the cables. It was found that such a system could accomplish crop management tasks, but certain limitations were also encountered that need to be overcome. These limitations included a constrained payload due to the weight of the platform causing the cables to sag, the additional labor and infrastructure needed to tension the cables, and the need for an improved method for stable electrical connectivity between the cables and the robot vehicle. Future applications for this work can include other crop management tasks which require high-power use for long durations, such as the UV-C treatment of plant fungal disease or laser weeding. Additionally, manipulators can be mounted on the platform for tasks such as pruning and harvesting. The cable drive system can potentially be adapted to other row crops as well, as long as the rows have regularly spaced support structures in place, such as in vineyards, orchards, and low-tunnel vegetable crops.