Environment scenario identification based on GNSS recordings for agricultural tractors.

• Defined five scenarios: open-sky, low-density trees, high-density trees, high wall/viaduct and tunnel/depot. • Environment scenario model built using a four-state Markov chain model. • A two-stage iterative matching algorithm for road segmentation and scenario identification. • Environment scenarios identification with an overall accuracy of 97.08%. A Global Navigation Satellite System (GNSS) is an essential part of the navigation system of autonomous tractor, which performs well in an open field but it might be affected by the local surroundings (e.g., tree canopies, buildings and overpasses) when traveling along a road. A reliable and cost-effective navigation system is important to ensuring that autonomous tractors operate reliably and safely with high precision positioning in a complex environments. The system requires an adaptive navigation strategy based on environment scenarios, and the identification of the environment scenario is a prerequisite. The present paper proposes an environment scenario identification model, laying a foundation for designing a scenario-based navigation strategy. First, the complex road surrounding environments are classified into five scenarios of open sky, low-density trees, high-density trees, high wall/overpass and tunnel/depot according to the GNSS positioning performance. Second, the environment is reconstructed using a four-state Markov chain model, and a two-stage iterative matching algorithm is designed to segment the recorded GNSS data and identify the scenario of each segment. The performance of the model is then evaluated using GNSS data sets recorded along the route from a depot to a field in Miyun, Beijing. The scenario identification results show that the developed model can identify all defined scenarios, with an overall accuracy of 97.08%. The identification accuracies for the scenarios of open sky, low-density trees, high-density trees, high wall/overpass and tunnel/depot are 97.32%, 96.30%, 95.35%, 94.29% and 100.00%, respectively. [ABSTRACT FROM AUTHOR]