Luigi Palmieri

I am a group leader and research scientist at Robert Bosch GmbH Corporate Research, where I work on artificial intelligence methods that enable autonomous agents, such as autonomous driving cars and robots, to learn and generate complex behaviors.

I earned my PhD with focus on robot planning, control and learning at Uni Freiburg, Social Robotics Laboratory. I obtained my Master and Bachelor degrees in computer engineering at the Università degli Studi della Campania, Luigi Vanvitelli.

Research Interests

My mission is to drive societal and economic progress by developing Embodied AI solutions that are intelligent, safe, and inherently socially aware. I believe that grounding AI in the physical world is the key to creating systems that seamlessly assist humans in their daily lives. My work focuses on the intersection of multi-modal perception and complex decision-making—creating “brains” for autonomous agents that can navigate the nuances of human interaction.

In my role as Group Manager, I am dedicated to leading talented teams to translate this vision into reality. I focus on developing innovative AI frameworks for single and multi-agent systems (across different domains, e.g. robotics, automated driving solutions, IoT), prioritizing socially-aware motion prediction, Vision-Language Models, and hybrid learning-planning architectures. By combining strategic leadership with technical excellence, I aim to deliver AI technologies that solve real-world challenges and improve our daily lives.

Academic Service

I’m a co-organizer of the Long-term Human Motion Prediction series of workshops, organizer and associate editor of the Special Issue on Long-term Human Motion Prediction on the IEEE Robotics and Automation Letters, served as associate editor for ICRA and PC/reviewer for: ICRA, IROS, RA-L, TRO, AAAI, and various ML/AI workshops.

news

Nov 14, 2025	Engaging discussions at the TU Delft ME Focus Day exploring how AI and robotics are shaping the future of manufacturing. Had wonderful conversations and gained valuable insights into the cutting-edge research happening at TU Delft.
Jun 10, 2025	We are finalizing our DARKO project. We had a great month of integration of our components and successful stakeholder demos at KI Fabrik Munich: a great team effort!
May 23, 2025	I had opportunity to present our recent works on context-aware robot navigation and foundation models at the workshops on Foundation Models and Neuro-Symbolic AI and Advances in Social Navigation at ICRA 2025.

selected publications

Delta: Decomposed efficient long-term robot task planning using large language models

Liu, Yuchen, Palmieri, Luigi, Koch, Sebastian, Georgievski, Ilche, and Aiello, Marco

In 2025 IEEE International Conference on Robotics and Automation (ICRA) 2025

Abs PDF

Long-term task planning is essential for mobile robots to achieve high-level goals in complex environments. Recent advances in large language models (LLMs) have shown promise in generating coherent and context-aware plans for robots. However, LLMs often struggle with long-term planning due to their limited context window and lack of domain-specific knowledge. In this paper, we present Delta, a novel framework that decomposes long-term task planning into manageable sub-tasks using LLMs. Delta leverages a hierarchical approach, where high-level goals are broken down into sub-goals, which are then further decomposed into executable actions. This decomposition allows Delta to effectively utilize LLMs for planning while mitigating their limitations. We evaluate Delta in a series of simulated and real-world experiments, demonstrating its ability to generate efficient and effective long-term plans for mobile robots. Our results show that Delta outperforms existing approaches in terms of plan quality, execution success rate, and adaptability to dynamic environments.
Grapheqa: Using 3d semantic scene graphs for real-time embodied question answering

Saxena, Saumya, Buchanan, Blake, Paxton, Chris, Chen, Bingqing, Vaskevicius, Narunas, Palmieri, Luigi, Francis, Jonathan, and Kroemer, Oliver

arXiv preprint arXiv:2412.14480 2025

Abs PDF

Embodied question answering (EQA) is a task that requires an agent to navigate and interact with its environment to answer a question. In this paper, we present Grapheqa, a novel approach that leverages 3D semantic scene graphs to enhance the agent’s understanding of its surroundings and improve its question-answering capabilities. Our method constructs a dynamic 3D semantic scene graph that captures the spatial relationships and attributes of objects in the environment. This graph is then used to inform the agent’s navigation and interaction strategies, allowing it to efficiently gather information relevant to the posed question. We evaluate Grapheqa on a benchmark EQA dataset and demonstrate that our approach outperforms existing methods in terms of accuracy and efficiency. Our results highlight the importance of incorporating structured environmental representations for effective embodied question answering.
Human motion trajectory prediction: A survey

Rudenko, Andrey, Palmieri, Luigi, Herman, Michael, Kitani, Kris M, Gavrila, Dariu M, and Arras, Kai O

The International Journal of Robotics Research 2020

Abs PDF

With growing numbers of intelligent autonomous systems in human environments, the ability of such systems to perceive, understand and anticipate human behavior becomes increasingly important. Specifically, predicting future positions of dynamic agents and planning considering such predictions are key tasks for self-driving vehicles, service robots and advanced surveillance systems. This paper provides a survey of human motion trajectory prediction. We review, analyze and structure a large selection of work from different communities and propose a taxonomy that categorizes existing methods based on the motion modeling approach and level of contextual information used. We provide an overview of the existing datasets and performance metrics. We discuss limitations of the state of the art and outline directions for further research.