Abstract

In order to build autonomous systems that can replace operator actions in agricultural operations, computer-based sensors and actuators, including global positioning systems, machine vision, and laser-based sensors, have been gradually incorporated into mobile robots. However, adding several electronic systems to a robot reduces its dependability and raises its price. For robotic systems to be practical, hardware minimization, software minimization, and simplicity of integration are crucial. The use of fleets of robots, in which several specialized robots work together to complete one or more agricultural tasks, is a development in the application of automation equipment in agriculture. In order to increase dependability, reduce complexity and costs, and enable the integration of software from various developers, this study aims to build a system architecture for both individual robots and robots working in fleets. There are several alternatives being researched, ranging from a totally dispersed design to a fully integrated architecture where a single computer manages all activities. This work adds more potential topologies while studying various topologies for managing robot fleets. Three ground mobile units built on a commercial tractor chassis make up the RHEA fleet, which is successfully using the architecture described in this study.

1. Initialization

The integration of numerous autonomous vehicles, particularly agricultural robots, has been made possible over the past 20 years thanks to specialized sensors (machine vision, GPS, real-time kinematics (RTK), laser-based equipment, and inertial devices), actuators (hydraulic cylinder, linear, and rotational electrical motors), and electronic equipment (embedded computers, industrial PCs, and PLCs) [1–5]. If given the appropriate implements (agricultural tools or utensils), these autonomous or semiautonomous systems can carry out precision agricultural operations since they are capable of exact placement and direction in the working field. The same kinds of sensors and actuators used in autonomous cars (GPS, machine vision, range finders, etc.) are also being employed to automate those tools (varying application rates of fertilizers or sprays, mechanical intrarow weed control, and seed planters) [6–11].

As a result, numerous sensors and/or actuators are duplicated when integrating a specific vehicle and a specific implement. What's more, a central, external computer must be utilized to manage the configuration of the vehicle and implement it. For agricultural machinery to be commercialized in a reliable, efficient, and cost-competitive manner, the vehicle implementation system's hardware must be kept to a minimum [12]. Therefore, creating a straightforward controller for both the truck and the implement would boost dependability, efficiency, and competitiveness.Many research teams are working on specific autonomous farming applications that will be used in the upcoming years [13–15], but many more are also seeking to manage a fleet of vehicles under a single controller. This is the emerging idea of robot fleets, which is a development for agriculture.

The theoretical underpinnings of robot fleets have lately been studied [16, 17], but the first agricultural applications are still in the works [18, 19]. The idea of eliminating redundant devices that coordinate various, heterogeneous systems by employing a single, external computer is prominent for this goal.Fleets of robots can offer a number of benefits [20–23]: employing a group of robots collaborating with one another to accomplish a specific goal is a developing and essential notion to realize the use of autonomous systems in routine agricultural operations. For high-value crops, where intelligent robots may replace time-consuming, expensive repetitive labor, the adoption of complicated and expensive systems will be appealing.

However, processing a large amount of data and managing numerous actuation signals are required for a robotic agricultural application, which might have a variety of technological challenges. Due to the fact that a failure in one robot component puts the entire fleet out of commission, an important limitation is that the number of total devices (such as sensors, actuators, and computers/controllers) increases in direct proportion to the number of fleet units. As a result, the mean time between failures drastically decreases. Fleet dependability, which is crucial for the application of automated systems to actual jobs and, in particular, to agriculture, is greatly impacted by this decrease in the time between failures.

The system architecture (consisting of sensors, actuators, and the computers performing the algorithms) for both the vehicle navigation system and the operation of the implement must be robust, simple, and modular in order to achieve a fleet of autonomous mobile robots for agricultural tasks that is flexible, reliable, and maintainable. The choice of the quantity and kind of sensors, actuators, and computers is one of the most crucial steps in the design of a control setup. Because the processes of perceiving and acting cannot be avoided, these components serve as the foundation for the design of the architecture and are very difficult to reduce in number. However, these sensors and actuators are typically managed by independent controllers, specifically commercial off-the-shelf (COTS) sensors like LIDARs and vision systems. Computers, however, are sufficiently adaptable.

The system, the crop field, and external elements, such as human supervisors, must all be operated concurrently in completely autonomous agricultural systems to ensure both efficiency and safety. A fully autonomous agricultural system is made up of these specific actions, which collectively include site-specific applications, autonomous navigation or remote operation, obstacle and interesting element detection, communication with external users or other autonomous units, and absolute or relative localization in the field. The autonomous vehicle and the autonomous implement are the two key components of this system (see Figure 1). The autonomous vehicle directs the agricultural system in a crop field to carry out a crop activity (for example, planting a crop), such as a customized commercial tractor, specialized platform, or small vehicle.

The autonomous implement will carry out these tasks (harvesting, hoeing, and weed management). Given the difficulty of the work, several specialized sensors and actuators are needed to complete the assignment in the environment.

In-depth research efforts have been reported in the literature for each individual system shown in Figure 1, with the goal of solving the autonomous guiding problem and the autonomous crop operation problem separately. The autonomous navigation problem is shown in Table 1 with a few instances, and the autonomous crop operation problem is presented in Table 2. Both tables include information about the application for which they were built as well as the primary sensor system that was employed.