Designing a perception system for safe autonomous operations in agriculture

In agriculture fields, autonomous machines employ high-tech systems and sensors and have been able to navigate automatically in certain fields without human intervention. Safety and reliability are key considerations in autonomous operations. A high-level perception is necessary to reduce human intervention or to even completely eliminate the human operator. In this paper, current perception technologies, normative references (ISO 25119 series, ISO 18497) and scientific literature were analyzed to prove the safety and reliability of autonomous driving under the complex surrounding. The main objective is to design a perception system that can reach at the least the lowest performance level for the reliability of safety. Several principles of tasks in relation to the stage in the life cycle of the system and hazardous situations and risks were examined. Analysis of different sensor modalities was performed, and sensor types were selected.