Autonomous vehicles (AV) are sophisticated systems comprising various sensors, powerful processors, and complex data processing algorithms that navigate autonomously to their respective goals. Out of several functions performed by an AV, one of the most important is developing situational intelligence to predict collision-free future trajectories. As an AV operates in environments consisting of various entities, such as other AVs, human-driven vehicles, and static obstacles, developing situational intelligence will require a collaborative approach. The recent developments in artificial intelligence (AI) and deep learning (DL) relating to AVs have shown that DL-based models can take advantage of information sharing and collaboration to develop such intelligence. However, most of these developments address only the requirements of urban environments, which are structured, and ignore the more challenging requirements of off-road environments, which are unstructured due to the lack of lane markings, traffic rules, and traffic signs. Given this deficiency and the lack of off-road vehicle motion and interaction data, we first employ two groups of AVs that will navigate in an unstructured environment to generate an off-road vehicle dataset. Based on the dataset, an encoder-decoder social long short-term memory (LSTM) network is developed to function as a situational intelligence model. The model comprises two data streams - a social interaction stream that learns how AVs interacts with their surroundings and a vehicle behavior stream that learns the dynamic characteristics of AVs. The model’s output is a multi-modal Gaussian distribution of all collision-free future AV trajectories. Finally, the effectiveness of the model is tested and verified using an independent subset of the dataset.
DISTRIBUTION A. Approved for public release; distribution unlimited. OPSEC#6871 (Pending, NOT approved for release).