Autonomous systems are increasingly used for military tasks. Military autonomous systems, and specifically land-based systems, require a good understanding of the terrain to be able to navigate a complex and fast-changing environment. It is therefore essential to understand the state of the art in modelling and reasoning on terrain elements for autonomous systems in a defence context. In this article we present a first review of research that addresses this topic. The collected literature can be divided into three process stages: Sensing, World Modelling and Reasoning. The state of the art is described in a compilation of 37 sources of literature divided into the mentioned process stages. In total we identified 7 articles on Sensing, 20 on World Modelling, 16 on Reasoning. We observed that many military applications focus on the aerial domain, including the manoeuvring and optimisation of UAVs for military tasks. Heuristic methods such as Monte Carlo methods and evolutionary computing are often employed and applied to the vehicle routing problem. These approaches have also been extensively applied to land warfare research, but have only been applied sparsely to sea warfare. The analysis of the 37 articles reveals a clear preference for heuristic methods, followed by Vehicle Routing Problem (VRP) techniques and machine learning approaches. Less frequently employed methods are graph algorithms, imitation learning, and reinforcement learning.