An agricultural robot that adjusts its pesticide spraying to suit the weather and terrain using a vision-language model

By cross-referencing crop type, the product used and weather data, the system adjusts the application rate and travel speed, detects rows with 30 per cent greater accuracy and minimises drift.

Precision spraying depends on many factors: temperature, wind, nature of the pesticide, and the robot's ability to locate plants. Current approaches focus on crop detection and apply fixed settings, where a human operator adapts his or her treatment to the conditions of the moment. Cong-Thanh Vu and Yen-Chen Liu propose a vision-language model framework that reasons on the scene and integrates multiple sources of information to decide how to spray.

Decision and steering

The model combines the type of crop, the product to be applied and the weather data to choose the spraying parameters; a predictive controller such as MPPI then follows the trajectory to precisely reach each plant.

According to the authors, row detection is at least 30% more accurate than compared to the methods compared, and tests in two environments show that the robot modulates its flow rate and speed while reducing product drift, which is both an economic and environmental issue.