LeaPRiDE: Learning, Planning, and Reasoning in
Dynamic Environments

Autonomous robots face significant challenges in dynamic environments - where robots operate alongside humans or interact with other robots. Factors such as uncertainty, partial observability, and unexpected disturbances in dynamic environments can severely limit the effectiveness of traditional planning and control framework. The ever-changing nature poses a great challenge to learning-based algorithms, as building a comprehensive dataset or simulator is nearly impossible. In addition, interaction with human teammates or multiple robots introduces additional complexity, including intent inference, safety constraints, and context-aware behavior. Dynamic environments further emphasize the increasing demands on the real-time reactivity of the planner, the smoothness of the generated trajectories, and the coherency of the decision. Together, these demands make the design of robust, scalable autonomous robots a painstakingly difficult challenge.

The workshop, LeaPRiDE, aims to bring researchers and practitioners from the human-robot interaction, multi-agent interaction, and robot learning communities to explore the latest advances, identify key challenges, share common concepts, and forge collaborations for autonomous robots operating in dynamic environments. By covering this broad spectrum of topics, LeaPRiDE seeks to inspire new research directions and foster synergy across domains, ultimately paving the way for more capable, adaptable, and trustworthy robotic systems.

Full-Day IROS Workshop

Date: 20 Oct. 2025, 9:00-17:00, Location: ROOM 102A

Submission: OpenReview

Submission Deadline: September 28, 23:59 AoE

Decision Deadline: October 10, 23:59 AoE

Topic

The LeaPRiDE workshop will cover a broad spectrum of research topics encountered in dynamic environments, including but not limited to:

- Learning frameworks for developing robust and adaptable policies in ever-changing environments.
- Data augmentation techniques to increase the diversity and realism of training datasets for dynamic scenarios.
- Simulations to generate varied and complex dynamic environments for training and evaluation.
- Fast and reactive motion planning algorithms for real-time adaptation to environmental changes and disturbances.
- Techniques for smooth, agile, and safe robot motions under high-speed and unpredictable conditions.
- Task and behavior abstractions for scalable high-level planning in dynamic, multi-agent environments.
- Hierarchical policy architectures for combining high-level decision-making with reactive low-level control.
- Intention inference methods for anticipating human actions or coordinating with other robots in collaborative tasks.
- World models to forecast environmental changes and agent behaviors for proactive decision-making.
- Frameworks for integrating Large Language Models (LLMs) into robotic systems while preserving low-latency, reactive control.
- Reasoning methods to improve explainability, interpretability, and robustness in unforeseen scenarios.
- Benchmarks and standardized tasks for evaluating robustness, reactivity, and safety in dynamic environments.

Invited Speakers

Dana Kulic

Professor

Monash University

Pulkit Agrawal

Associate Professor

MIT

David Hsu

Chair Professor

National University of Singapore

Markus Wulfmeier

Researcher
Google Deepmind

Bin He

Professor
Tongji Univeristy

Yoonchang Sung

Assistant Professor
Nanyang Technological University

Schedule

A detailed schedule will be announced after acceptance.

Organizers

Puze Liu

Researcher
DFKI

Guanya Shi

Assist. Professor
CMU

Davide Tateo

Group Leader
TU Darmstadt

Christopher Mower

Researcher
Huawei R&D

Ruolin Ye

PhD student
Cornell Univeristy

Advisory Board

Jan Peters

TU Darmstadt

Haitham Bou Ammar

Huawei & UCL

Tapomayukh Bhattacharjee

Cornell Univeristy

For any questions regarding the workshop organization, please contact:
Puze Liu (puze.liu--at--dfki--dot--de)

Page updated

Google Sites

Report abuse

LeaPRiDE: Learning, Planning, and Reasoning in Dynamic Environments