Bienemann, Alexander (Universität der Bundeswehr Munich), Mortimer, Peter (Universität der Bundeswehr München), Luettel, Thorsten (Universität der Bundeswehr München), Maehlisch, Mirko (University of German Military Forces Munich)

Surface-Aware Lane Departure for Obstacle Avoidance in Off-Road Environments Using Semantic Segmentation and MPC

Scheduled for presentation during the Regular Session "S26b-Motion Planning, Trajectory Optimization, and Control for Autonomous Vehicles" (TH-EA-T26), Thursday, November 20, 2025, 13:50−14:10, Broadbeach 1&2

2025 IEEE 28th International Conference on Intelligent Transportation Systems (ITSC), November 18-21, 2025, Gold Coast, Australia

This information is tentative and subject to change. Compiled on October 18, 2025

Abstract

When driving autonomously on public roads, the vehicle is usually not allowed to leave the roadsides. However, when driving off-road, smooth terrain and unpaved paths often resemble a lane that the vehicle can drive in and leave to avoid an obstacle blocking the way if the surface next to the lane allows it. In this paper, we propose a new method for using semantic segmentation to inform a model predictive controller (MPC) about the surface of the surrounding area. In this way, the MPC can actively decide whether to leave the lane to avoid an obstacle or not. The MPC even decides if it is better to leave the lane to the left or right. We validate the performance of our approach through practical experiments on a real autonomous vehicle, showcasing that the system is capable of working in the real world and in real-time. To get additional information about lane boundaries and obstacles, we also use an occupancy grid map and a multi-object tracker.