A fully autonomous solar-powered lightweight weeding robot, using AI for plant recognition, precision contact and contactless weeding methods suited for hard soils, hilly terrains and arid climates.

Project Information

RoboAIweeder

Grant agreement ID: 101166300

DOI

10.3030/101166300

EC signature date 26 May 2024

Start date 1 June 2024

End date 30 November 2026

Funded under

• The European Innovation Council (EIC)

Total cost

€ 3 372 767,00

EU contribution

€ 2 360 936,90

Coordinated by

SMART FARM ROBOTIKS
Bulgaria

Periodic Reporting for period 1 - RoboAIweeder (A fully autonomous solar-powered lightweight weeding robot, using AI for plant recognition, precision contact and contactless weeding methods suited for hard soils, hilly terrains and arid climates.)

Reporting period: 2024-06-01 to 2025-05-31

Summary of the context and overall objectives of the project

The proposed project centres on accelerating the R&D activities and go-to-market of the precision weeding robot being developed of our company, Smart Farm Robotix. The robot aims to address in a paradigm-changing way the fight against weeds, one of the oldest problems faced by farmers everywhere. Weed pressure is a very serious problem for farmers, decreasing yields and worsening produce quality, which is also aggravated by recent human-activity-induced trends like emergence of herbicide-resistant weeds and spread of hard-to-contain invasive plant species. Current weeding solutions like herbicides, machine tilling and manual weeding each have serious drawbacks and are becoming ever more untenable due to multiple societal and environmental trends.
Our solution to this problem is a fully autonomous weeding robot in the form of a light-weighted 4-wheeled electric engines rover, using 100% solar energy to power all onboard systems. A down-facing camera captures images of plants, which are then analysed by Artificial Intelligence based on deep neural networks to distinguish weeds from cultivated plants and guide the weeding modules. The weeds are removed via contact (mechanical) or contactless (energy beam) weeding modules, depending on weed size, type and soil conditions. Off-the-shelf multi-GNSS (including Galileo) navigation system, combined with both cameras inputs and support software, provides end-of-row turning capability and robust special orientation around the field. Several integrated devices ensure constant 2-way communication with a central company data and monitoring hub.
The combination of multiple cutting-edge technologies (AI, robotics, photonics, mechanics, long-range and satellite communication) into a single operating robot, will allow us to offer farmers a combination of precision, effectiveness, scalability and reliability totally unachievable by current weeding methods.

Work performed from the beginning of the project to the end of the period covered by the report and main results achieved so far

Work package 2 – Robotic Platform
Added active breaks to wheels, Improved suspension and strengthened robot main frame, Validated active suspension options, Full energy system data tracking and monitoring (software module developed), Improved solar panels introduced – lighter and more efficient
Work package 3 – Weeding Systems
Optimized the 3-arm gripper for improved performance and stability. 2x increase of weeding speed, Several prototypes are being tested in the field, Continued improvement of speed (already 2x) , stability and precision
Work package 4 – Plants Recognition & Analysis
Constant increase in field plants images processed for improved NNs recognition accuracy , Improvements to the aiming system in various challenging conditions – dust on plants, pebbles, straws in the field, Extensive use of field annotation saving time for back-end operations, Added 2 new crops – coriander and peppers, Introduced efficient and rapid turn-around images processing – filming, labelling, AI-feeding, re-deployment in field
Reviewed initial IR images from current cameras for first tell-tale signs of plants stress
Work package 5 – Robot Self-navigation and Security
Started experiments with front cameras for improved in-row navigation, Started exploring algorithms on field images multiple rows together (e.g. as in peppers), Started integrating GPS readings for better while-working field mappin, Successfully integrated and tested various European LTE providers modules, making the robots ready for international wor, Started developing an app for smoother tablet-robot communication

Progress beyond the state of the art and expected potential impact (including the socio-economic impact and the wider societal implications of the project so far)

The project aims to combine in an innovative way several high-end and cutting-edge technologies, but above all, develop new ones crucial to meeting the needs and the specific requirements of our target customers.
First, the weeding robot makes use of some of the latest research in computer image recognition and the field of artificial intelligence, which is only recently beginning to be implemented in agriculture on a commercial scale. We take that research to the next level by developing it in a commercially viable product, which can recognize multiple plants species in various weather conditions 24/7. The image recognition process is also aided significantly by using some of the latest image processing units available, in particular the Nvidia Jetson line of products. These devices are energy efficient and NN-enabled optimized combinations of CPU and GPU , and targeted exactly for autonomous devices using large numbers of images for navigation and other purposes.
Second, we are developing a fully autonomous platform which can reliably self-navigate in open fields. This is a ground-breaking innovation which will eliminate the need for constant human supervision and presence in the field during the weeding process, present in both currently used manual and machine weeding methods. We achieve it by combining high-tech sensors like latest multi-GNSS receivers (including Galileo), cameras and others, but also enhancing them as a system, by developing innovative algorithms which can make use of all these sensors readings combined and provide robust and effective self-navigation in and around the farm fields.
Third, we are making use of some of the latest available photonics devices, which we combine in a working system designed for our specific weeding needs. It also uses a system of optical instruments and is mounted on a 3D robotic arm, which thanks to advanced algorithms can be positioned, aimed and applied towards weeds, destroying them without physical contact. What’s more, we achieve this without materially raising the plant’s temperature, thus avoiding potential fire hazard in the fields, unique among all similar start-up projects. This allows us both to save people the need to work in the open field in dangerous environments , but also to operate safely in hot arid climates.

picture1.jpg

picture2.png

picture6.png

picture5.jpg

picture4.jpg

picture3.jpg