The second prototype of the harvesting robot from Floating Robotics is much more precise, moving faster between the rows of tomatoes this fall than last year. According to its creator, Salman Faraji, the device is set to replace 3.5 workers in the future.
Many Swiss market gardeners struggle with major staff fluctuations, even in the middle of the season. This means that in the morning, there might not be enough people for the tomato harvest, so others have to work overtime.
Machines are more reliable - if they work. This is exactly what Salman Faraji from Floating Robotics, a start-up from Zurich that emerged from an ETH spin-off and is working on developing a harvesting robot, is working on.
Important findings from the first prototype
Last year, the engineer had an initial prototype with a robotic arm reach for cherry tomatoes in two greenhouses. The robot primarily consisted of cameras for tomato recognition, software and a robotic arm. Harvesting and defoliation already worked quite well, but there was still a lack of precision and speed.
However, first tests were inconclusive. Salman Faraji, however, drew valuable insights from the initial trials and, together with his technical partner Mohsen Moosavi, set to work in the following months, optimizing the entire machine.
Faster gripper arm and better camera
The new prototype was ready in September of this year: a new robot arm is now permanently mounted on a converted trolley, the cameras have been extended and a new packaging unit with a separate unloading trolley has been constructed. The new gripper arm with the knives now works four times faster than last year, says Salman Faraji.
However, the focus of this year's tests was on precision. "The biggest challenge is for the cameras to reliably recognize the fruit stems of the tomato panicles and guide the robot arm with the scissors and gripper to the right place."
This is primarily a matter of deep learning: the software is fed thousands of images - in this case of tomato plants - and an algorithm is created with the help of which the robot can ultimately make the right decisions. The new camera technology can now correctly detect 90 percent of the panicles, says the robotics expert. Compared to the first prototype, panicles that grow one behind the other are now being recognized better as well.
Market launch planned for 2025
As last year, vegetable growing company Beerstecher AG is making its greenhouse in Hinwil available for the tests. Salman Faraji's contact there is greenhouse manager Bianca Curcio. "If we want new solutions for the future, we have to support projects like this," she says, explaining their commitment.
She is convinced that this type of automation will become a reality. It is therefore interesting for the company to follow the development of the harvesting robot on site and to be able to control it to some extent. During the trials this fall, she noticed big improvements compared to the previous year. "The robot has become faster and more precise." She is certain that Salman Faraji will gain more valuable insights and she is already looking forward to its implementation next year.
Salman Faraji will continue to develop the robot over the next few months, including night work and automatic row changing. His goal is for one picking robot to cover 1.5 hectares of greenhouse area in the future, replacing 3.5 workers.
He is currently still looking for more funding. But the interest shown from all over the world gives him a lot of confidence. Marketing of the first units is planned for 2025.
Source: www.lid.ch
For more information:
ETH Zürich - Robotic Systems Lab (RSL)
Tel.: +41 44 633 84 93
[email protected]
www.floatingrobotics.com