arXiv subjects

Se Hwan Jeon

Publications and source records attributed to Se Hwan Jeon.

10 recordsLinked to original sources

arXiv2026

WarpMPC: Large-Batch MPC on GPU via ADMM with Unrolled $LDL^\top$ Factorization

This paper introduces numerical optimizations for maximizing throughput on GPU when solving large batches (10,000 to over 100,000) of sequential quadratic programming (SQP) iterations, where all problems have the same structure. The optimizations are implemented in a toolbox WarpMPC for model-predictive control (MPC) in JAX and Warp. Based on the insight that all MPC problem instances in a batch share the same sparsity in time, cost, and constraints, we propose unrolling sparse linear factorizations and solves, which dominate alternating direction method of multipliers (ADMM) solver runtime. We avoid memory access bottlenecks and wasting computations via optimized memory layout, padding-reducing segmentation of the unrolled factorization, and dependency level scheduled backsolves, additionally accelerating sensitivity computation. We achieve throughputs of 8,000 to 250,000 SQP iterations per second on nonlinear cartpole, quadrotor, and humanoid robot benchmarks, outperforming baselines by 3$\times$ to 25$\times$. We illustrate practical usefulness by synthesizing a dataset and training a neural network approximation of an MPC in under 4 minutes that stabilizes a nano quadrotor in hardware experiments.

cs.RO↗

arXiv2026

Learning Reactive Dexterous Grasping via Hierarchical Task-Space RL Planning and Joint-Space QP Control

In this work, we propose a hybrid hierarchical control framework for reactive dexterous grasping that explicitly decouples high-level spatial intent from low-level joint execution. We introduce a multi-agent reinforcement learning architecture, specialized into distinct arm and hand agents, that acts as a high-level planner by generating desired task-space velocity commands. These commands are then processed by a GPU-parallelized quadratic programming controller, which translates them into feasible joint velocities while strictly enforcing kinematic limits and collision avoidance. This structural isolation not only accelerates training convergence but also strictly enforces hardware safety. Furthermore, the architecture unlocks zero-shot steerability, allowing system operators to dynamically adjust safety margins and avoid dynamic obstacles without retraining the policy. We extensively validate the proposed framework through a rigorous simulation-to-reality pipeline. Real-world hardware experiments on a 7-DoF arm equipped with a 20-DoF anthropomorphic hand demonstrate highly robust zero-shot transferability for dexterous grasping to a diverse set of unseen objects, highlighting the system's ability to reactively recover from unexpected physical disturbances in unstructured environments.

cs.RO↗

arXiv2025

Residual MPC: Blending Reinforcement Learning with GPU-Parallelized Model Predictive Control

Model Predictive Control (MPC) provides interpretable, tunable locomotion controllers grounded in physical models, but its robustness depends on frequent replanning and is limited by model mismatch and real-time computational constraints. Reinforcement Learning (RL), by contrast, can produce highly robust behaviors through stochastic training but often lacks interpretability, suffers from out-of-distribution failures, and requires intensive reward engineering. This work presents a GPU-parallelized residual architecture that tightly integrates MPC and RL by blending their outputs at the torque-control level. We develop a kinodynamic whole-body MPC formulation evaluated across thousands of agents in parallel at 100 Hz for RL training. The residual policy learns to make targeted corrections to the MPC outputs, combining the interpretability and constraint handling of model-based control with the adaptability of RL. The model-based control prior acts as a strong bias, initializing and guiding the policy towards desirable behavior with a simple set of rewards. Compared to standalone MPC or end-to-end RL, our approach achieves higher sample efficiency, converges to greater asymptotic rewards, expands the range of trackable velocity commands, and enables zero-shot adaptation to unseen gaits and uneven terrain.

cs.RO↗

arXiv2025

Speech to Reality: On-Demand Production using Natural Language, 3D Generative AI, and Discrete Robotic Assembly

We present a system that transforms speech into physical objects using 3D generative AI and discrete robotic assembly. By leveraging natural language, the system makes design and manufacturing more accessible to people without expertise in 3D modeling or robotic programming. While generative AI models can produce a wide range of 3D meshes, AI-generated meshes are not directly suitable for robotic assembly or account for fabrication constraints. To address this, we contribute a workflow that integrates natural language, 3D generative AI, geometric processing, and discrete robotic assembly. The system discretizes the AI-generated geometry and modifies it to meet fabrication constraints such as component count, overhangs, and connectivity to ensure feasible physical assembly. The results are demonstrated through the assembly of various objects, ranging from chairs to shelves, which are prompted via speech and realized within 5 minutes using a robotic arm.

cs.RO↗

arXiv2025

Making Physical Objects with Generative AI and Robotic Assembly: Considering Fabrication Constraints, Sustainability, Time, Functionality, and Accessibility

3D generative AI enables rapid and accessible creation of 3D models from text or image inputs. However, translating these outputs into physical objects remains a challenge due to the constraints in the physical world. Recent studies have focused on improving the capabilities of 3D generative AI to produce fabricable outputs, with 3D printing as the main fabrication method. However, this workshop paper calls for a broader perspective by considering how fabrication methods align with the capabilities of 3D generative AI. As a case study, we present a novel system using discrete robotic assembly and 3D generative AI to make physical objects. Through this work, we identified five key aspects to consider in a physical making process based on the capabilities of 3D generative AI. 1) Fabrication Constraints: Current text-to-3D models can generate a wide range of 3D designs, requiring fabrication methods that can adapt to the variability of generative AI outputs. 2) Time: While generative AI can generate 3D models in seconds, fabricating physical objects can take hours or even days. Faster production could enable a closer iterative design loop between humans and AI in the making process. 3) Sustainability: Although text-to-3D models can generate thousands of models in the digital world, extending this capability to the real world would be resource-intensive, unsustainable and irresponsible. 4) Functionality: Unlike digital outputs from 3D generative AI models, the fabrication method plays a crucial role in the usability of physical objects. 5) Accessibility: While generative AI simplifies 3D model creation, the need for fabrication equipment can limit participation, making AI-assisted creation less inclusive. These five key aspects provide a framework for assessing how well a physical making process aligns with the capabilities of 3D generative AI and values in the world.

cs.RO↗

arXiv2025

Learning Humanoid Arm Motion via Centroidal Momentum Regularized Multi-Agent Reinforcement Learning

Humans naturally swing their arms during locomotion to regulate whole-body dynamics, reduce angular momentum, and help maintain balance. Inspired by this principle, we present a limb-level multi-agent reinforcement learning (RL) framework that enables coordinated whole-body control of humanoid robots through emergent arm motion. Our approach employs separate actor-critic structures for the arms and legs, trained with centralized critics but decentralized actors that share only base states and centroidal angular momentum (CAM) observations, allowing each agent to specialize in task-relevant behaviors through modular reward design. The arm agent guided by CAM tracking and damping rewards promotes arm motions that reduce overall angular momentum and vertical ground reaction moments, contributing to improved balance during locomotion or under external perturbations. Comparative studies with single-agent and alternative multi-agent baselines further validate the effectiveness of our approach. Finally, we deploy the learned policy on a humanoid platform, achieving robust performance across diverse locomotion tasks, including flat-ground walking, rough terrain traversal, and stair climbing.

cs.RO↗

arXiv2024

CusADi: A GPU Parallelization Framework for Symbolic Expressions and Optimal Control

The parallelism afforded by GPUs presents significant advantages in training controllers through reinforcement learning (RL). However, integrating model-based optimization into this process remains challenging due to the complexity of formulating and solving optimization problems across thousands of instances. In this work, we present CusADi, an extension of the CasADi symbolic framework to support the parallelization of arbitrary closed-form expressions on GPUs with CUDA. We also formulate a closed-form approximation for solving general optimal control problems, enabling large-scale parallelization and evaluation of MPC controllers. Our results show a ten-fold speedup relative to similar MPC implementation on the CPU, and we demonstrate the use of CusADi for various applications, including parallel simulation, parameter sweeps, and policy training.

cs.RO↗

arXiv2024

Learning Emergent Gaits with Decentralized Phase Oscillators: on the role of Observations, Rewards, and Feedback

We present a minimal phase oscillator model for learning quadrupedal locomotion. Each of the four oscillators is coupled only to itself and its corresponding leg through local feedback of the ground reaction force, which can be interpreted as an observer feedback gain. We interpret the oscillator itself as a latent contact state-estimator. Through a systematic ablation study, we show that the combination of phase observations, simple phase-based rewards, and the local feedback dynamics induces policies that exhibit emergent gait preferences, while using a reduced set of simple rewards, and without prescribing a specific gait. The code is open-source, and a video synopsis available at https://youtu.be/1NKQ0rSV3jU.

cs.RO↗

arXiv2023

Benchmarking Potential Based Rewards for Learning Humanoid Locomotion

The main challenge in developing effective reinforcement learning (RL) pipelines is often the design and tuning the reward functions. Well-designed shaping reward can lead to significantly faster learning. Naively formulated rewards, however, can conflict with the desired behavior and result in overfitting or even erratic performance if not properly tuned. In theory, the broad class of potential based reward shaping (PBRS) can help guide the learning process without affecting the optimal policy. Although several studies have explored the use of potential based reward shaping to accelerate learning convergence, most have been limited to grid-worlds and low-dimensional systems, and RL in robotics has predominantly relied on standard forms of reward shaping. In this paper, we benchmark standard forms of shaping with PBRS for a humanoid robot. We find that in this high-dimensional system, PBRS has only marginal benefits in convergence speed. However, the PBRS reward terms are significantly more robust to scaling than typical reward shaping approaches, and thus easier to tune.

cs.RO↗

arXiv2021

Real-time Optimal Landing Control of the MIT Mini Cheetah

Quadrupedal landing is a complex process involving large impacts, elaborate contact transitions, and is a crucial recovery behavior observed in many biological animals. This work presents a real-time, optimal landing controller that is free of pre-specified contact schedules. The controller determines optimal touchdown postures and reaction force profiles and is able to recover from a variety of falling configurations. The quadrupedal platform used, the MIT Mini Cheetah, recovered safely from drops of up to 8 m in simulation, as well as from a range of orientations and planar velocities. The controller is also tested on hardware, successfully recovering from drops of up to 2 m.

cs.RO↗