A Safety Architecture for Autonomous Agricultural Vehicles

Abstract

Sixty years after debuting in industrial environments, robots are making their way into our everyday life. Farmers have benefited for some time from self-guided machinery including combines and harvesters. More recently, multi-purpose autonomous vehicles have started to be deployed in orchards, groves, nurseries, and other agricultural environments to automate or augment operations such as pruning, thinning, harvesting, mowing, and spraying. Successful commercialization of such vehicles will depend heavily on them being able to operate safely and avoid accidents involving humans, animals, trees, and farm infrastructure. We propose a safety architecture to guide the design and deployment of autonomous agricultural vehicles and their introduction into production environments. The architecture spans the three elements that, combined, should ensure safe operation over a wide spectrum of applications: (1) a distributed, sensor-based, intelligent decision-making system that coordinates and guides fleets of vehicles in and around orchards and other agricultural environments; (2) multimodal interfaces for workers to interact with the vehicles using natural language, gestures, and portable devices; (3) a comprehensive regulatory framework of standards for vehicle safety that covers everything from basic robotic technology to advanced behaviors. In this paper we present the fundamental aspects of this agricultural robotic safety architecture. We illustrate its application with examples of autonomous agricultural vehicles we developed in the past that lay down the path toward full introduction of safe, intelligent machines in agricultural production environments.