Marlon Steiner

Marlon Steiner

PhD student in the field of motion planning for autonomous driving

Publications

You can also find my articles on my Google Scholar profile.

MAP-Former: Multi-Agent-Pair Gaussian Joint Prediction

Published in preprint on arxiv.org (soon available: IEEE Intelligent Vehicles Symposium (IV 2024)), 2024

There is a gap in risk assessment of trajectories between the trajectory information coming from a traffic motion prediction module and what is actually needed. Closing this gap necessitates advancements in prediction beyond current practices. Existing prediction models yield joint predictions of agents´ future trajectories with uncertainty weights or marginal Gaussian probability density functions (PDFs) for single agents. Although, these methods achieve high accurate trajectory predictions, they only provide little or no information about the dependencies of interacting agents. Since traffic is a process of highly interdependent agents, whose actions directly influence their mutual behavior, the existing methods are not sufficient to reliably assess the risk of future trajectories. This paper addresses that gap by introducing a novel approach to motion prediction, focusing on predicting agent-pair covariance matrices in a “scene-centric” manner, which can then be used to model Gaussian joint PDFs for all agent-pairs in a scene. We propose a model capable of predicting those agent-pair covariance matrices, leveraging an enhanced awareness of interactions. Utilizing the prediction results of our model, this work forms the foundation for comprehensive risk assessment with statistically based methods for analyzing agents´ relations by their joint PDFs.

Recommended citation: Steiner, M., Klemp, M. & Stiller, C. (2024). "MAP-Former: Multi-Agent-Pair Gaussian Joint Prediction." arXiv preprint arXiv:2404.19283. https://arxiv.org/abs/2404.19283

SceneMotion: From Agent-centric Embeddings to Scene-wide Forecasts

Published in (soon) IEEE 27th International Conference on Intelligent Transportation Systems (ITSC), 2024

Self-driving vehicles rely on multimodal motion forecasts to effectively interact with their environment and plan safe maneuvers. We introduce SceneMotion, an attention-based model for forecasting scene-wide motion modes of multiple traffic agents. Our model transforms local agent-centric embeddings into scene-wide forecasts using a novel latent context module. Our latent context module learns a scene-wide latent space from multiple agent-centric embeddings, enabling joint predictions and interaction modeling. The competitive performance in the Waymo Open Interaction Prediction Challenge demonstrates the effectiveness of our approach. Furthermore, we use our scene-wide forecasts to identify areas where future interaction between agents is likely, enriching the contextual information for motion planners.

Recommended citation: Wagner, R., Taş, Ö., Steiner, M., Konstantinidis, F., Königshof, H., Klemp, M., Fernandez, C. & Stiller, C. (2024). "SceneMotion: From Agent-centric Embeddings to Scene-wide Forecasts." IEEE 27th International Conference on Intelligent Transportation Systems (ITSC).

AUTOtech.agil: Architecture and Technologies for Orchestrating Automotive Agility

Published in Aachener Kolloquium Fahrzeug- und Motorentechnik GbR, 2023

Future mobility will be electrified, connected and automated. This opens completely new possibilities for mobility concepts that have the chance to improve not only the quality of life but also road safety for everyone. To achieve this, a transformation of the transportation system as we know it today is necessary. The UNICARagil project, which ran from 2018 to 2023, has produced architectures for driverless vehicles that were demonstrated in four full-scale automated vehicle prototypes for different applications. The AUTOtech.agil project builds upon these results and extends the system boundaries from the vehicles to include the whole intelligent transport system (ITS) comprising, e.g., roadside units, coordinating instances and cloud backends. The consortium was extended mainly by industry partners, including OEMs and tier 1 suppliers with the goal to synchronize the concepts developed in the university-driven UNICARagil project with the automotive industry. Three significant use cases of future mobility motivate the consortium to develop a vision for a Cooperative Intelligent Transport System (C-ITS), in which entities are highly connected and continually learning. The proposed software ecosystem is the foundation for the complex software engineering task that is required to realize such a system. Embedded in this ecosystem, a modular kit of robust service-oriented modules along the effect chain of vehicle automation as well as cooperative and collective functions are developed. The modules shall be deployed in a service-oriented E/E platform. In AUTOtech.agil, standardized interfaces and development tools for such platforms are developed. Additionally, the project focuses on continuous uncertainty consideration expressed as quality vectors. A consistent safety and security concept shall pave the way for the homologation of the researched ITS.

Recommended citation: van Kempen, R. et al. (2023). "AUTOtech.agil: Architecture and Technologies for Orchestrating Automotive Agility." 32nd Aachen Colloquium Sustainable Mobility 2023. https://publications.rwth-aachen.de/record/971700

Safety Reinforced Model Predictive Control (SRMPC): Improving MPC with Reinforcement Learning for Motion Planning in Autonomous Driving

Published in IEEE 26th International Conference on Intelligent Transportation Systems (ITSC), 2023

Model predictive control (MPC) is widely used for motion planning, particularly in autonomous driving. Real-time capability of the planner requires utilizing convex approximation of optimal control problems (OCPs) for the planner. However, such approximations confine the solution to a subspace, which might not contain the global optimum. To address this, we propose using safe reinforcement learning (SRL) to obtain a new and safe reference trajectory within MPC. By employing a learning-based approach, the MPC can explore solutions beyond the close neighborhood of the previous one, potentially finding global optima. We incorporate constrained reinforcement learning (CRL) to ensure safety in automated driving, using a handcrafted energy function-based safety index as the constraint objective to model safe and unsafe regions. Our approach utilizes a state-dependent Lagrangian multiplier, learned concurrently with the safe policy, to solve the CRL problem. Through experimentation in a highway scenario, we demonstrate the superiority of our approach over both MPC and SRL in terms of safety and performance measures.

Recommended citation: Fischer, J., Steiner, M., Taş, Ö. & Stiller, C. (2023). "Safety Reinforced Model Predictive Control (SRMPC): Improving MPC with Reinforcement Learning for Motion Planning in Autonomous Driving." IEEE 26th International Conference on Intelligent Transportation Systems (ITSC). https://doi.org/10.1109/ITSC57777.2023.10422605