RoPL

Example Event

Sat, 01 Jun 2030 13:00:00 +0000

Slides can be added in a few ways:

Create slides using Wowchemy’s Slides feature and link using slides parameter in the front matter of the talk file
Upload an existing slide deck to static/ and link using url_slides parameter in the front matter of the talk file
Embed your slides (e.g. Google Slides) or presentation video on this page using shortcodes.

Further event details, including page elements such as image galleries, can be added to the body of this page.

Mimic Intent, Not Just Trajectories

Sat, 07 Feb 2026 00:00:00 +0000

Vec-QMDP: Vectorized POMDP Planning on CPUs for Real-Time Autonomous Driving

Fri, 06 Feb 2026 00:00:00 +0000

Any House Any Task (AHAT) is Now Public

Sun, 01 Feb 2026 00:00:00 +0000

🚀 Exciting News! We are thrilled to announce the release of our latest work:“Any House Any Task: Scalable Long-Horizon Planning for Abstract Human Tasks.” This research tackles long-horizon planning in large environments given ambiguous human instructions.

Paper Details

Title: Any House Any Task: Scalable Long-Horizon Planning for Abstract Human Tasks

Authors: Zhihong Liu, Yang Li, Renming Huang, Cewu Lu, Panpan Cai

Abstract: Open world language conditioned task planning is crucial for robots operating in large-scale household environments. While many recent works attempt to address this problem using Large Language Models (LLMs) via prompting or training, a key challenge remains scalability. Performance often degrades rapidly with increasing environment size, plan length, instruction ambiguity, and constraint complexity. In this work, we propose Any House Any Task (AHAT), a household task planner optimized for long-horizon planning in large environments given ambiguous human instructions. At its core, AHAT utilizes an LLM trained to map task instructions and textual scene graphs into grounded subgoals defined in the Planning Domain Definition Language (PDDL). These subgoals are subsequently solved to generate feasible and optimal long-horizon plans through explicit symbolic reasoning. To enhance the model’s ability to decompose complex and ambiguous intentions, we introduce TGPO, a novel reinforcement learning algorithm that integrates external correction of intermediate reasoning traces into Group Relative Policy Optimization (GRPO). Experiments demonstrate that AHAT achieves significant performance gains over state-of-the-art prompting, planning, and learning methods, particularly in human-style household tasks characterized by brief instructions but requiring complex execution plans.

Paper: Available on arXiv
Project Page: AHAT Project Website

We are excited to share this work with the robotics community and look forward to your feedback and potential collaborations!

Any House Any Task: Scalable Long-Horizon Planning for Abstract Human Tasks

Sun, 01 Feb 2026 00:00:00 +0000

Hi-Drive: Hierarchical POMDP Planning for Safe Autonomous Driving in Diverse Urban Environments

Thu, 09 Oct 2025 00:00:00 +0000

Hi-Drive Paper Accepted by IEEE Robotics and Automation Letters (RA-L)

Fri, 26 Sep 2025 00:00:00 +0000

🎊 Excellent news! We are delighted to announce that our paper “Hi-Drive: Hierarchical POMDP Planning for Safe Autonomous Driving in Diverse Urban Environments” has been accepted for publication in IEEE Robotics and Automation Letters (RA-L).

Paper Details

Title: Hi-Drive: Hierarchical POMDP Planning for Safe Autonomous Driving in Diverse Urban Environments

Authors: Xuanjin Jin, Chendong Zeng, Shengfa Zhu, Chunxiao Liu, Panpan Cai

Abstract: Uncertainties in dynamic road environments pose significant challenges for behavior and trajectory planning in autonomous driving. This paper introduces Hi-Drive, a hierarchical planning algorithm addressing uncertainties at both behavior and trajectory levels using a hierarchical Partially Observable Markov Decision Process (POMDP) formulation. Hi-Drive employs driver models to represent uncertain behavioral intentions of other vehicles and uses their parameters to infer hidden driving styles. By treating driver models as high-level decision-making actions, our approach effectively manages the exponential complexity inherent in POMDPs. To further enhance safety and robustness, Hi-Drive integrates a trajectory optimization based on importance sampling, refining trajectories using a comprehensive analysis of critical agents. Evaluations on real-world urban driving datasets demonstrate that Hi-Drive significantly outperforms state-of-the-art planning-based and learning-based methods across diverse urban driving situations in real-world benchmarks.

IEEE Robotics and Automation Letters (RA-L) is a premier publication venue for robotics research, known for its rapid review process and high-quality contributions to the field of robotics and automation. This work represents a significant contribution to autonomous driving research, particularly in addressing the complex challenges of planning under uncertainty in urban environments.

Congratulations to all the authors on this outstanding achievement!

Two Papers Accepted by NeurIPS 2025

Fri, 26 Sep 2025 00:00:00 +0000

🎉 Congratulations to our research team! We are thrilled to announce that two papers from our group have been accepted by the Conference on Neural Information Processing Systems (NeurIPS) 2025.

Featured Work

Our two accepted papers are:

Paper 1: TRU-POMDP

Paper: Available on arXiv
Project Page: TRU-POMDP Project Website

Paper 2: UniDomain

Paper: Available on arXiv
Project Page: UniDomain Project Website

We look forward to presenting our work and sharing our findings with the broader research community at NeurIPS 2025. Stay tuned for more details about the papers and presentations as we approach the conference!

From Lab to Cafe: Boba Bot Puts Our [NeurIPS'25] UniDomain Research to the Ultimate Test

Fri, 12 Sep 2025 00:00:00 +0000

At the recent Anniversary Open Day of the Shanghai Innovation Institute (SII), our research came to life with a simple request: “Hey BobaBot, make me a bubble tea!”

Our robotic barista, “Boba Bot,” was put to the test, running non-stop for two consecutive days. It seamlessly managed the entire workflow from order-taking to task-planning and final execution, demonstrating remarkable resilience under pressure. This was more than a demo; it was a validation of our work in a live, unpredictable setting, far from the ideal conditions of a lab.

Boba Bot serves as the official real-world showcase for our [NeurIPS'25] UniDomain project. The ability of the robot to operate autonomously and consistently is powered by the core principles of our UniDomain framework, bridging the gap between theoretical research and practical, tangible application.

Contact

Tue, 06 May 2025 00:00:00 +0000

Pretraining a Unified PDDL Domain from Real-World Demonstrations for Generalizable Robot Task Planning

Wed, 01 Jan 2025 00:00:00 +0000

Tru-POMDP: Task Planning Under Uncertainty via Tree of Hypotheses and Open-Ended POMDPs

Wed, 01 Jan 2025 00:00:00 +0000

World Modeling Makes a Better Planner: Dual Preference Optimization for Embodied Task Planning

Wed, 01 Jan 2025 00:00:00 +0000

Multi-agent reinforcement learning for autonomous driving: A survey

Mon, 01 Jan 2024 00:00:00 +0000

Neural randomized planning for whole body robot motion

Mon, 01 Jan 2024 00:00:00 +0000

Rethinking State Disentanglement in Causal Reinforcement Learning

Mon, 01 Jan 2024 00:00:00 +0000

What truly matters in trajectory prediction for autonomous driving?

Sun, 01 Jan 2023 00:00:00 +0000

People

Mon, 24 Oct 2022 00:00:00 +0000

Tour

Mon, 24 Oct 2022 00:00:00 +0000

Closing the planning--learning loop with application to autonomous driving

Sat, 01 Jan 2022 00:00:00 +0000

GAMMA: A general agent motion model for autonomous driving

Sat, 01 Jan 2022 00:00:00 +0000

LEADER: Learning Attention over Driving Behaviors for Planning under Uncertainty

Sat, 01 Jan 2022 00:00:00 +0000

MAGIC: Learning Macro-Actions for Online POMDP Planning using Generator-Critic

Wed, 01 Jan 2020 00:00:00 +0000

Summit: A simulator for urban driving in massive mixed traffic

Wed, 01 Jan 2020 00:00:00 +0000

Context and intention aware planning for urban driving

Tue, 01 Jan 2019 00:00:00 +0000

LeTS-Drive: Driving in a Crowd by Learning from Tree Search

Tue, 01 Jan 2019 00:00:00 +0000

HyP-DESPOT: A hybrid parallel algorithm for online planning under uncertainty

Mon, 01 Jan 2018 00:00:00 +0000

Point cloud based path planning for tower crane lifting

Mon, 01 Jan 2018 00:00:00 +0000

Porca: Modeling and planning for autonomous driving among many pedestrians

Mon, 01 Jan 2018 00:00:00 +0000

Automatic path planning for dual-crane lifting in complex environments using a prioritized multiobjective PGA

Sun, 01 Jan 2017 00:00:00 +0000

Parallel genetic algorithm based automatic path planning for crane lifting in complex environments

Fri, 01 Jan 2016 00:00:00 +0000

Simulation-enabled vocational training for heavy crane operations

Fri, 01 Jan 2016 00:00:00 +0000

A GPU-enabled parallel genetic algorithm for path planning of robotic operators

Wed, 01 Jan 2014 00:00:00 +0000

Collision detection using axis aligned bounding boxes

Tue, 01 Jan 2013 00:00:00 +0000

A New Approach to Design Rational Harmonic Surface over Rectangular or Triangular Domain

Sat, 01 Jan 2011 00:00:00 +0000

Mon, 01 Jan 0001 00:00:00 +0000