We are completing autonomous navigation in the vineyard using a visual system

December 21, 2020  |  Development

The Androver II robotic test platform, completely developed by RoboTech Vision, can move autonomously in a vineyard thanks to our algorithm. It mainly uses an RTV sensor Box developed by RoboTech Vision. The robot can thus navigate between the rows of the vineyard, go from row to row, as well as detect and avoid possible obstacles without the help of a human.

RoboTech Vision strives to develop universal algorithms that can be easily applied to a wide variety of devices. The company is therefore testing them on several platforms, such as the Husky A200 robot with a differential chassis or their own Androver II robot with a Double Ackerman chassis. Each type of robot uses a different type of autonomous navigation, depending on its hardware capabilities.

After almost two years of developing autonomous navigation algorithms, the company has managed to develop an algorithm using artificial intelligence that allows the robot to move autonomously in the vineyard. It is a reactive navigation without the use of GPS, where the robot responds to the environment. It uses the IMU and especially the RTV sensor Box with its visual system.

How does it work?

The RTV sensor Box is a universal device that combines a 360° camera, a front camera and a 3D laser scanner. “The robot uses the forward and 360° camera with the help of a neural network to distinguish between the grapevines and foreign objects. With the help of LiDAR, the robot orients itself in the vineyard, avoids obstacles and stays in the middle between the rows,” says Ing. Matej Vargovčík, a robotics engineer from RoboTech Vision, who developed autonomous navigation in the vineyard.

Before starting the application, the user selects how many lines the robot should drive autonomously. The robot starts from the starting position to which it is navigated by the teleoperator or arrives at the place without assistance. “The RTV sensor Box can be used not only in the vineyard, but also for autonomous navigation on the road. The idea is to combine these two types of navigation so that the robot could autonomously get to the starting point in the vineyard,” adds engineer Vargovčík.

The difference between these two navigations is that, unlike the road, the vineyard does not need to be mapped in advance. “While navigation on the road with an RTV sensor Box uses GPS and a map in addition to the visual system, navigation in the vineyard is based only on the visual system. The RTV sensor Box thus fulfills the function of the robot’s eyes,” explains Vargovčík.

Safety comes first

Once the robot is in the right place, it will enter the first row autonomously using our algorithms. Live view from the robot’s camera and its telemetry data can be monitored from the control center thanks to WiFi network. If necessary, the user can take control of the robot with a joystick. The algorithm also includes the detection of obstacles that the robot is trying to avoid or stop in front of them. The device also includes a stop button that disconnects the robot’s drives and stops it immediately.

In addition to autonomous row navigation, the robot can move to the next row without human assistance. If a device on which the algorithm is applied has a large turning radius, it can be set to passage through even rows and return through odd rows. Autonomous navigation is also adapted for a vineyard whose rows do not end in one parallel line.

The algorithm can also be deployed on existing devices if it contains the necessary sensor and hardware equipment. Autonomous navigation in the vineyard can thus be used, for example, for mowing grass between rows, spraying the vineyard or cultivating the soil around the roots of the vineyard. The algorithm can also be applied to orchards where trees are planted in rows.

Accuracy and precision

According to professor Ing. Pavel Pavloušek, Ph.D. from Mendel University in Brno, the use of robots in viticulture has proven to be very important. “For several years, I have been working on the use of unmanned monitoring of vineyards to improve the identification of grapevine infestation by pathogens and the prognosis of fungal diseases. Experience so far has shown that it would be appropriate to extend aerial surveillance by monitoring certain parameters directly in the vineyard. Robots can provide an invaluable service in this regard,” he claims.

According to Pavloušek, viticulture is also struggling with a shortage of manual workers, and therefore every effort is made to mechanize all work processes in the vineyard. “The advantage of robots can be accuracy and precision. It would therefore be appropriate to use them in vineyards for work assignments where the quality of the operation can make a significant contribution to greening viticulture,“ thinks the professor.

According to the professor, it is also very important to minimize the mechanical damage to the plants. “Damage can increase the susceptibility to biotic stress, such as the complex of vine trunk disease, the most well-known disease which is known as ESCA in our country,“ says the professor. „I therefore think that drones and robots can become a common part of viticulture in the near future,“ he adds.