Remote Operation of a Harvesting Robot via Edge Computing-Based Quality Control
-- Contributing to the advancement of user-friendly smart agriculture --

News Highlights:

Background

Japanese agriculture is currently facing a serious labor shortage due to a declining and aging workforce. In response, the adoption of smart agriculture technologies is accelerating.
　Until now, smart agriculture has mainly focused on stabilizing crop yields by optimizing the timing and amount of watering and fertilization. However, much of the fieldwork still relies heavily on manual labor. As a result, there is a growing need to reduce the workload through the use of agricultural robots.

In particular, plant-based agriculture poses unique challenges for automation, as tasks such as harvesting and pesticide application often require nuanced human judgment. This makes the use of remotely operated robots, which can incorporate human decision-making, especially important.
　In this demonstration, we envisioned farm support workers performing harvesting operations remotely and took on the challenge of verifying the operability of a remote-controlled harvesting robot.

To make remote operation practical, it is essential not only to reduce network and server latency but also to visualize otherwise invisible factors—such as network delays—through image processing to support smooth operation and accurate harvesting decisions.
　Furthermore, in scenarios where heavy system loads cause unavoidable latency increases, mechanisms that maintain operability without performance degradation are key to the widespread adoption of remote operation technologies.

Against this backdrop, we conducted a proof-of-concept aimed at enabling remote farm support workers to perform smooth and efficient harvesting by using agricultural robots equipped with AI-powered image processing.

Overview and Results of the Demonstration Experiment

In this demonstration, we established a remote harvesting environment in which an operator in Tokyo controlled a robot located in a strawberry field in Akita Prefecture, using video transmitted via a network. A camera mounted on the harvesting robot captured video footage, which was analyzed by an image processing server to determine whether each strawberry was ready for harvest. This judgment was overlaid on the video feed. The operator used this enhanced video displayed on a monitor to remotely control the robot and harvest strawberries in Akita.

The system configuration included a harvesting robot and image processing server installed in the strawberry field in Akita, and a control terminal located at a site in Tokyo. The two locations, approximately 400 km apart in a straight line, were connected via fiber optic access and internet lines (Figure 1).
　Under these conditions, we confirmed that the robot could be remotely operated with high precision, enabling strawberries to be harvested without damage.

Additionally, we implemented a feature in the system that adjusts both the monitor display and the robot arm's speed according to end-to-end network delay. We then evaluated how much this feature improved operability during strawberry harvesting.
　Specifically, we measured the success rate—defined as the percentage of times the robot arm could be moved accurately to a strawberry's position in a single attempt. In environments with fluctuating delays, the success rate was approximately 50%. By utilizing this feature, the success rate improved to about 80%, representing a 30% increase.
　Furthermore, all five test participants reported that they perceived a clear improvement in operability due to the new feature.

Figure 1 Configuration of Remote Harvesting System

Key Technologies

The high-speed closed-loop control technology for network and computing² enables real-time monitoring of both processing time at the edge server and network delay. When the end-to-end delay exceeds performance requirements, the system instantly switches to an alternative route and server process, maintaining a stable, low-latency service. In this demonstration, the technology was enhanced and adapted for use in smart agriculture applications.

Figure 2 System Integration Function

Figure 3 System Integration Function

Roles of Each Company

Future Developments

The results of this demonstration highlight the advantages of improved operability for agricultural robots under fluctuating network conditions. We will continue working toward the practical application of these findings to help realize a new form of agriculture in Japan that integrates human expertise with robotic capabilities.

Furthermore, the high-speed closed-loop control technology for network and computing is not limited to agriculture—it offers solutions to labor and technical skill shortages faced across various industries. We will continue to promote the practical deployment and broader adoption of this technology to enable new, location-independent ways of working and to enhance labor productivity, ultimately contributing to the resolution of societal challenges in the industrial sector.

[Glossary]

¹Tsukuba Forum 2025
https://www.rd.ntt/e/as/tforum/

²High-speed closed-loop control technology for network and computing
https://group.ntt/en/newsrelease/2023/05/16/230516a.html

About NTT

NTT contributes to a sustainable society through the power of innovation. We are a leading global technology company providing services to consumers and businesses as a mobile operator, infrastructure, networks, applications, and consulting provider. Our offerings include digital business consulting, managed application services, workplace and cloud solutions, data center and edge computing, all supported by our deep global industry expertise. We are over $90B in revenue and 340,000 employees, with $3B in annual R&D investments. Our operations span across 80+ countries and regions, allowing us to serve clients in over 190 of them. We serve over 75% of Fortune Global 100 companies, thousands of other enterprise and government clients and millions of consumers.