TRUST-AD

Accepted Papers

Automatic Curriculum Learning for Driving Scenarios: Towards Robust and Efficient Reinforcement Learning

Ahmed Abouelazm, Tim Weinstein, Tim Joseph, Philip Schörner, Marius Zöllner

This paper addresses the challenges of training end-to-end autonomous driving agents using Reinforcement Learning (RL). RL agents are typically trained in a fixed set of scenarios and nominal behavior of surrounding road users in simulations, limiting their generalization and real-life deployment. While Domain Randomization offers a potential solution by randomly sampling driving scenarios, it frequently results in inefficient training and sub-optimal policies due to the high variance among training scenarios. To address these limitations, we propose an automatic curriculum learning framework that dynamically generates driving scenarios with adaptive complexity based on the agent’s evolving capabilities. Unlike manually designed curricula that introduce expert bias and lack scalability, our framework incorporates a ”teacher” that automatically generates and mutates driving scenarios based on their learning potential—an agent-centric metric derived from the agent’s current policy, eliminating the need for expert design. The framework enhances training efficiency by excluding scenarios the agent has mastered or finds too challenging. We evaluate our framework in a reinforcement learning setting where the agent learns a driving policy from camera images. Comparative results against baseline methods, including fixed scenario training and domain randomization, demonstrate that our approach leads to enhanced generaliza- tion, achieving higher success rates, +9% in low traffic density, +21% in high traffic density, and faster convergence with fewer training steps. Our findings highlight the potential of ACL in improving the robustness and efficiency of RL-based autonomous driving agents

A Quasi-Steady-State Black Box Simulation Approach for the Generation of G-G-G-V Diagrams

Frederik Werner, Simon Sagmeister, Mattia Piccinini, Johannes Betz

The classical g-g diagram, representing the achievable acceleration space for a vehicle, is commonly used as a constraint in trajectory planning and control due to its computational simplicity. To address non-planar road geometries, this concept can be extended to incorporate g-g constraints as a function of vehicle speed and vertical acceleration, commonly referred to as g-g-g-v diagrams. However, the estimation of g-g-g-v diagrams is an open problem. Existing simulation-based approaches struggle to isolate non-transient, open-loop stable states across all combinations of speed and acceleration, while optimization-based methods often require simplified vehicle equations and have potential convergence issues. In this paper, we present a novel, open-source, quasi-steady-state black box simulation approach that applies a virtual inertial force in the longitudinal direction. The method emulates the load conditions associated with a specified longitudinal acceleration while maintaining constant vehicle speed, enabling open-loop steering ramps in a purely QSS manner. Appropriate regulation of the ramp steer rate inherently mitigates transient vehicle dynamics when determining the maximum feasible lateral acceleration. Moreover, treating the vehicle model as a black box eliminates model mismatch issues, allowing the use of high-fidelity or proprietary vehicle dynamics models typically unsuited for optimization approaches. An open-source version of the proposed method is available at: https://github.com/TUM-AVS/GGGVDiagrams

Safer and Trustworthier Navigation of Automated Vehicles

Martin Aleksandrov

We propose a novel approach for the safer and trustworthier navigation of automated vehicles, which uses risk estimations for predicting vehicle trajectories and generates navigation instructions for these trajectories. To increase the reliability of instructions, we first use the latest YOLO11 model and adapt images from the KITTI dataset to include rain, fog, and evening effects, such as low lighting and high darkness, and, then, we train the model on the modified images to make it more robust to scene changes due to such weather and evening conditions. For this more robust model, we give functions that assign risk estimations to images taken from cameras associated with nodes in road networks. Based on these risk estimations, we give algorithms for the dynamic navigation of automated vehicles along network trajectories of low risk. In cases where risk estimations at network nodes and in network paths are too high, we give methods that produce human-interpretable explanations of these risk values and recommend driving instructions. Overall, our research generally facilitates the road integration and social adoption of automated driving.