Autonomous perception and hoeing of crops in sustainable agriculture – PEAD

One of the main priorities of sustainable agriculture is the reduction of crop inputs, to promote production agriculture that respects the environment and the health of the consumer. The PEAD project aims to develop a mechanical and autonomous crop hoeing solution, that aims to reduce by 100% the input of herbicides associated with crops. It meets the reduction targets for plant protection production and deals with two major points in the operational research and development programs of the Ecophyto II plan:
- the proposal of an innovative solution for hoeing, based on autonomous robotics and integrated perception of the environment in complex situations (in terms of mobility, observation and detection conditions),
- the proposal of a persistent and long-term mapping of crop and flora evolutions, in scope of their sustainable management.

The proposed solution in PEAD project covers the three facets of the ROSE Challenge call:
- to improve observation and detection, by providing solutions that are based on deep learning for detection and differentiation of cultivated plants amongst the weeds;
- to analyze/decide, with a reliable and robust decision-making process, based on learning and convolutional neural networks;
- to act and operate in a targeted way, by hoeing systematically the weeds and protecting the plants of crops.

The PEAD project aims to develop an autonomous system of perception, mobility and intervention in hoeing, targeting the adventitious flora in crops at different stages of development and in different environmental conditions (meteorology conditions, day/night, and with frequencies adapted to seasonal and crop growth activities).