Improving Crop Yields Using the ROVY Platform for Robots
Robots to the Rescue - Solving Challenges in Agriculture
The agricultural sector faces many challenges today. The increased costs of raw materials, labor, and transportation have also naturally led to inflation in food costs. The effects of climate change forces many producers to adapt to a changing ecology. Consumers and policymakers around the globe are adding pressure to reduce the use of potentially harmful pesticides and herbicides, which are key tools used to improve crop yield.
Increasingly, robotic automation is being seen as an important way to solve some of all of these problems. They can help with selective weed removal, scan fields, assist with harvesting, and even help with timely and precision pollination.
To perform these tasks, robotic systems, as well as drones, require the right vision and rugged compute platform. In this article, we discuss how TechNexion’s products, including its embedded vision cameras and its latest system-on-module, ROVY-4VM, can be an end-to-end solution for robots and drones in agricultural automation. We will explore its features, ease-of-use, and how it can reduce costs and time-to-market for robot manufacturers.
How robots and drones can help with precision farming
Many agricultural practices lead to operational challenges in addition to unnecessary expenses. They lack precision and have limitations when it comes to scalability. By leveraging new-age technologies like cameras and intelligent processing systems, robots and drones can provide an important way to solve these challenges.
Weed Removal
Using high-resolution cameras and next-gen computing systems to find weeds among crops, robots can identify non-beneficial plants (weeds). Using several different techniques, weeds can be eliminated. One technique, developed by Carbon Robotics, uses a high energy laser to literally cook the weeds, and only the weeds, killing them in real time. Another method, used by John Deere in their “See and Spray” product, selectively applies a targeted herbicide to precisely inhibit weed growth, reducing herbicide use and costs. This can be done over a very long boom arm, which allows many rows to be treated at once, reducing the time and number of passes required.
Field Monitoring and Surveying
Aerial vehicles (Drones) can help farmers observe crop conditions from the air and take necessary steps to optimize crop yield. Drones leverage Near InfraRed cameras to capture images for measuring NDVI (Normalized Difference Vegetation Index) to determine the spread of vegetation and water in the field.
Some of the other functions of agricultural drones include topographic surveying, soil monitoring, spreading fertilizer, and collecting soil & water samples.
Optimal Pollination
According to bee experts at the Food and Agriculture Organization (FAO) of the United Nations, a third of the world’s food production depends on bees. However, the bee population is in decline. In the US alone, commercial honey bee colonies were down 7 percent from 2022 to 2023. This requires us to come up with newer ways of enabling pollination, and an important tool may be robotic systems.
Using pollination vehicles with mechanical arms, such as products from BloomX, pollen can be collected from plants and transported to another. Coupled with services that utilize advanced AI models, these devices can determine the best time for pollination by considering a wide variety of parameters – such as weather conditions, time of year, and crop species.
Robotic systems can further enhance this process through autonomous operation of the pollination vehicles, enabling both pollination operation and monitoring of crops to ensure precise timing for pollination. To do this, they require need state of the art compute systems that can perform all processing onboard and support all the necessary communication within the vehicle. High-quality camera modules are also essential. In the next section, we will look at the key requirements of robotic vision systems used for agricultural purposes.
Agricultural Robots: Key Requirements
Agricultural robots are highly demanding in terms of their performance. They make use of a complex sensor and computing system to implement an array of functions required for autonomous navigation, path planning, object and recognition, and 3D terrain mapping. With their complex architecture, agricultural robots should have the following capabilities:
- High-performance processing: The processing unit of the robot should be capable of handling and processing the data captured and delivered by its cameras and sensors and implement control loops in real time.
- Vision processing: Modern-day agricultural robots need advanced vision processing capabilities for performing video pre-processing, motion, stereo vision processing for depth perception, multi-camera support, video encoding and decoding.
- Accelerated AI/ML processing: Agricultural robots rely on AI and ML algorithms to perform various analyses. For example, robots may need to assess the ripeness of fruits and vegetables to determine if they are ready for plucking. Another example is an automated weeding robot identifying weeds based on characteristics such as species and size. These need accelerated AI processing power of several TOPS (Trillions or Tera Operations per Second).
- Robustness and industrial grade build: Agricultural robots must operate in harsh external weather conditions. The components in them may be subjected to shocks, vibrations, extreme temperature & pressure, etc. The system should be able to withstand such conditions, which in turn ensures the longevity and performance of the robot.
- Real time control: Robots need to act fast. For instance, consider an automated weeding robot that moves around an agricultural field. It must stop instantly when it sees an obstacle, a person, animal, is near a virtual fence, or is otherwise off-track. This means that the reaction time of robots has to be a few milliseconds and have little tolerance for missed deadlines. The control loop may operate on the order or 100Hz, or 1kHz.
- Safety: The safety system in a robot is responsible for ensuring that it is made ‘safe’ in the case of certain faults, regardless of the state of the main controller (e.g. the processor). It is responsible to take certain actions if a particular threshold is exceeded. For example, if the robot is getting too close to an obstacle or person, or even if there is a hardware or software fault, the safety system can take certain measures such as restricting motor speeds, limiting motor torque, and even disabling power to the drive system.
- Domain-applicable I/O: The core system unit of the robot should have domain-applicable connectivity options to ensure compatibility with external sensors and actuators while minimizing integration costs.
TechNexion’s ROVY Platform
TechNexion’s focus on innovation, and new product development that helps solve real-world challenges, has given rise to another series of products called ROVY. The ROVY platform is a robust system on module specially designed for robotic and industrial control applications. It is a highly integrated system that comes with all the necessary components needed in a robot.
Integration
The traditional IPC-based architecture of robots is comprised of different off-the-shelf assemblies like the controller, vision accelerators, industrial wireless routers and ethernet switches, and a safety PLC, as separate components as seen in the image below.
TechNexion’s ROVY platform provides all of the building blocks together into a single system on module as shown in the illustration below.
Within the ROVY System on Module are a:
- Sophisticated vision processing system
- High performance edge AI processing
- Ample general-purpose CPU and GPU
- Industry-leading I/O integration
- Realtime controllers
- Isolated MCU subsystem for safety processing
Let’s dive into each of the key features of the platform.
Scalable high-end video and AI compute capabilities
The ROVY system platform comes with an AI processing ability of up to 32 TOPS that will meet the demands of most modern autonomous mobile robots. It can support up to 12 camera inputs, making it suitable for multi-camera operation. In addition, the integrated vision processing engine makes it possible to do perform most video operations, including motion and stereo processing in dedicated hardware accelerators on the platform itself.
Right-sized I/O for robotics
One of the biggest highlights of the ROVY platform is that it comes inbuilt with all the I/O a robotic system needs, which significantly reduces the external complexity of a robotic system, removing the need for many external adapters to communicate with sensors and actuators. The platform has the following connectivity options:
- Multiple (up to 9) GbE, including 10 GbE support, using a built-in high-performance Ethernet switch
- Up to 4x PCIe Gen3 ports
- Up to 10 CANbus ports
- Up to 2x USB3.1
- Up to 2x MIPI-DSI ports for display
- Up to 3x SPI, 6x I2C, 6x UARTs, and many GPIO
Pre-integrated support for multiple TechNexion cameras
One of the advantages for robot manufacturers while using the ROVY platform is that it comes as a ready to use solution for TechNexion cameras. You don’t have to worry about finding a suitable camera for your robot that can work well with ROVY. Our MIPI cameras and FPD-Link III cameras can be readily integrated with ROVY, making the whole development process easy for you. This will also help you reduce your time to market significantly.
With these cutting-edge features, ROVY is perfect for robotic applications, particularly agricultural robots used for automated weed removal. It can also be used in drones that are used for tasks such as measuring NDVI, soil analysis, and spreading fertilizer. Overall, ROVY can help farmers improve crop yield and improve the effectiveness of precision farming practices.
ROVY-4VM – a TI TDA4VM system on module
The first among the ROVY series of products from TechNexion is ROVY-4VM. It is based on the TI TDA4VM Jacinto™ processor. It comes with 2x 64-bit Cortex A72 + 6x R5F MCUs for real time processing in addition to having built-in DSPs and ISP. It is capable of processing up to 8 camera streams simultaneously and performs AI processing of up to 8 TOPS.
Each ROVY-4VM SOM is a self-contained compute solution consisting of the Texas Instruments TDA4VM SOC, up to 8GB RAM, a power management system capable of independent supply of safety and main processing domains, and Octal SPI and UFS non-volatile storage.
To learn more about the features and benefits of ROVY-4VM, visit the product page.
Wrapping up
With ROVY and its wide portfolio of cameras, TechNexion brings to the table an end-to-end solution for robotic applications. Whether you are building a new robot or wish to switch from your existing IPC-based system, we can help. ROVY is an all-in-one package that will reduce your integration efforts while our cameras – with features such as high resolution, global shutter, Near Infrared optimization – ensure your robots capture high-quality images needed for vision and AI processing. Talk to our experts today to elevate your robotic product development game.
Related Products
Get a Quote
Fill out the details below and one of our representatives will contact you shortly.