Towards the responsible development of autonomous robotics in agriculture: a call to action
David Christian Rose, Jessica Lyon, Auvikki de Boon Marc Hanheide and Simon Pearson of the University of Reading, the Lincoln Institute, Agri-Food Technology, University of Lincoln and the Riseholme Campus.
Faced with immediate labour shortages, combined with existing threats such as climate change, biodiversity loss, and an increasing population, agriculture is in need of new ideas. Autonomous robots are one solution to the urgent global reality of reduced labour due to the COVID-19 pandemic (plus by political uncertainty in some places), and further offer the potential to sustainably intensify production by reducing carbon emissions, chemical use, and waste. They also offer a lifeline to struggling farm businesses who are being forced to consider down-scaling operations due to lack of labour. Despite the promise of automation, there are, however, a number of social, environmental, legal, and ethical issues to consider. Autonomous technologies are rapidly approaching high-technology readiness levels and will soon be available for on-farm implementation. However, we know that the history of agricultural technology revolutions is littered with examples of failure, particularly when new technologies have been undeveloped in a responsible way. In this perspective, we argue that proponents of autonomous robotics should re-balance activities to focus more on embracing responsible innovation principles. We note that there has been little, if any, empirical research that sets out to operationalize responsible innovation principles for the development of autonomous robots in farming and identify key areas for action. We discuss different anticipation exercises that can be carried out by innovators to foresee the consequences of their technology and evaluate different methods of inclusion so that a wide range of stakeholders can be involved in setting the future vision for robotics in agriculture. We further consider how developers can be reflexive, how they might adapt their robots so that they are safe, secure, and reliable, and also how institutional frameworks in industry and policy can support responsive and adaptable innovation systems to new knowledge. We provide examples of how responsible innovation principles can be embedded into the user-centered design of autonomous robots and identify key questions for the future research agenda in this vital but often side-lined area of scholarship.
Remote Monitoring Control and Security Technology for Full Self- Driving Agricultural Machinery
Yoshikazu Kusumi, Akira Nagai, Takuya Murayama, Fumiaki Kudoh and Masaru Ohzeki of Nippon Telegraph and Telephone and NTT Domoco.
In this presentation, we introduce two cutting-edge technologies that will be necessary for the future of agriculture. One of them is the technology needed for field-to-field self-driving and remote monitoring and control of robotic agricultural machinery. The other is a security technology that prevents agricultural robots from being manipulated by malicious actors.
Japanese agricultural industry has been facing a labor shortage due to the long-term decline in the number of farmers and the aging population. Additionally, the spread of COVID-19 this year has caused the number of foreign technical interns to decrease, and the agricultural workforce is shrinking more quickly. For the preservation and development of the Japanese agricultural industry, rapid productivity improvements that will expand the farmland of every farmer are needed. The maximum possible automation of agricultural work through the use of robotic agricultural machinery is also expected to resolve the labor shortage we are facing.
In this context, we have been conducting research and developing technologies to achieve the highest level of smart agriculture in the world. So far, we have achieved self-driving agricultural machinery and remote monitoring control towards Level 3 using cutting edge robotic agriculture technology, the fifth-generation mobile communication system (5G), multiple network optimization technology, which is one of the technologies that created the innovative network technology IOWN, and high-precision positioning technology. With this, we can achieve tables and smooth wide-area self-driving of agricultural machinery.
The more widespread the use of self-driving agricultural machines, the more important it will be to strengthen security to prevent unauthorized operation of robotic agricultural machines by cyber threat actors. We have developed lightweight authentication and authorization technology for agricultural robots based on cryptographic theory. Some agricultural machines have limited resources such as CPU and memory. In addition, the communication speed may become slow depending on the usage environment. Therefore, our technology is effective because the computational cost is small, and the communication volume is low. This technology can be applied not only to robotic agricultural machines, but also to devices such as sensors and other end devices as well as devices such as webcams, thus contributing to the entire agricultural sector.
We use these cutting-edge technologies to contribute to resolving the labor shortage and improving agricultural productivity.
Towards Context-Aware Navigation in Agricultural Environments For Long-Term Autonomous Robots
Benjamin Kisliuk, Mark Höllmann, Christoph Tieben, Jan Christoph Krause, Sebastian Pütz, Santiago Focke Martinez, Alexander Mock, Felix Igelbrink, Thomas Wiemann, Stefan Stiene and Joachim Hertzberg of the German Research Center for Artificial Intelligence (DFKI), Plan Based Robot Control, Osnabrück University and the Institute of Computer Science.
Currently, most robotic systems are remote controlled or are based on a single global environment representation. This fixed representation and the fixed set of behaviors limits the flexibility of autonomous robots. In agriculture, the environment changes fast during vegetation periods. This demands context-sensitive navigation. We want to enable our robotic system AROX (Autonomous Robotic Experimental Platform) to map fields over an entire vegetation period without any human intervention and to adapt its action schemes to the environment. For this purpose, we set up the hardware infrastructure for housing and charging the robot, and implemented the required software based on the Robot Operating System (ROS) to realize the desired context sensitivity.
For common static environments, especially indoor scenarios, established solutions, e.g. based on static grid maps, are available. But for long-term autonomous applications, it must be considered that a change in environment might be expectable in some regions. A change can also be unforeseen elsewhere. Thus, it is not possible to rely solely on the established solutions. Depending on the layouts of the fields, the condition of the plants, the season and the weather specific solutions are required. However, the established solutions are well tested and understood for simple applications and can therefore be used to solve subproblems.
We want to consider the semantic and geometric context by combining the classical algorithms with novel application-specific methods. For this purpose, we deployed the autonomous system AROX for plant monitoring, in a real field environment. The system consists of two main components: the mobile robot itself, which is able to drive around and sense the environment, and a container located next to the field, which contains the loading device and gives shelter to the robot. The aim in this example scenario is to enable the robot to create 3D scans of a cornfield on a regular basis without any user interaction.
We annotated a geometric map of the surroundings with semantic information and divided the map into different navigation zones. This allows us to choose the most appropriate planning and navigation method specialized for the specific semantic context. The context is indicated by the annotated geometric environment model. In this contribution we present the implementation of the technical aspects for robust navigation of such a system based on a semantic map. We introduce the system components of the used mapping system and the ROS-based implementation of context sensitive operation.
The presented AROX prototype is able to navigate safely through different zones within the environment. To further improve autonomy, an execution monitoring system is required for long-term autonomy to detect abnormal behavior of the system and unforeseen events. For this purpose, the current semantic environment map will be combined with a symbolic model to enable rule-based reasoning for process state detection. Such asemantic monitoring system would allow automatic recognition of process states on an abstract level, such as "loading", "driving" or "scanning", based on the sensor data and the semantic environment model. Furthermore, illegal, or abnormal states can also be detected by defining simple rules.
The DFKI Niedersachsen Lab (DFKINI) is sponsored by the Ministry of Science and Culture of Lower Saxony and the Volkswagen Stiftung.