Osiris⁺⁺: hierarchical representations for robotic-enabled precision agriculture

Area of Science:

• Agricultural Robotics
• Human-Robot Interaction
• Precision Agriculture

Background:

• Agricultural robotics development aims to boost efficiency and address labor shortages.
• Current human-robot interaction in agriculture requires technical expertise, limiting adoption, especially in Africa due to limited exposure and linguistic diversity.
• Existing collaboration methods are confined to structured indoor environments.

Purpose of the Study:

• To introduce Osiris++, a flexible system for seamless human-robot communication in precision agriculture.
• To enable natural language interaction, including African languages, for agricultural robots.
• To overcome adoption barriers in the agricultural sector.

Main Methods:

• Development of Osiris++, a novel approach for human-robot communication.
• Validation and evaluation of Osiris++ in real-world agricultural settings.
• Utilizing scene graphs to interpret and act on natural language instructions.

Main Results:

• Osiris++ successfully creates accurate and useful scene graphs in agricultural environments.
• The system demonstrates potential for natural language instruction processing.
• The approach facilitates improved human-robot collaboration for precision agriculture tasks.

Conclusions:

• Osiris++ offers a flexible solution for natural language-based human-robot communication in agriculture.
• The system has the potential to significantly increase the adoption of agricultural robotics, particularly in diverse linguistic contexts.
• This work paves the way for more intuitive and accessible human-robot collaboration in precision farming.