Kowndinya Boyalakuntla

Introduction

I am a third-year Ph.D. student in Computer Science at Rutgers University with nearly 4 years of experience in robot learning and planning. My research spans 3D scene understanding, object rearrangement, dynamic manipulation, and model-based reinforcement learning for humanoid locomotion. I have published at CoRL and IROS. My current research focuses on planning with learned world models for robust humanoid loco-manipulation.

Education

Research experience

Publications

First-Author

1. PL-MPC: Beyond Policy Alignment: Closing the Planning–Learning Loop for Robot Control with Learned World Models

   Kowndinya Boyalakuntla, Yuhan Liu, Abdeslam Boularias

   Under review · Manuscript, 2026

   [Paper] [Project] [Code]
   • Combined uncertainty-aware MPPI planning with value learning and policy distillation using learned world models; improved mean return over the TD-M(PC)² baseline on the challenging HumanoidBench tasks balance-hard (98 to 387) and hurdle (199 to 466), using Unitree H1 in MuJoCo.
   • Demonstrated zero-shot sim-to-real transfer of a learned world model coupled with an MPPI planner from Isaac Gym to a KUKA IIWA14 for tool manipulation on a challenging wrench–nut alignment task.

2. Plan-Conditioned Imitation for Robust Object Retrieval under Self-Occlusion in Dense Clutter

   Kowndinya Boyalakuntla, Ajinkya Pawar, Abdeslam Boularias, Jingjin Yu

   Under review · Manuscript, 2026

   [Paper] [Project] [Code] [Data]
   • Developed a plan-conditioned imitation policy that rearranges dense clutter to create grasping clearance even when the robot’s arm blocks the camera’s view, with potential applications in bin picking and assembly-kit retrieval.
   • Achieved 90.7% retrieval success in Isaac Gym simulation and 90% after sim-to-real transfer to UR5e, cutting hardware task time by 65% versus the full-observation planner.

3. KARL: Kalman-Filter Assisted Reinforcement Learner for Dynamic Object Tracking and Grasping

   Kowndinya Boyalakuntla, Abdeslam Boularias, Jingjin Yu

   IROS, 2025

   [Paper] [Code]
   • Combined curriculum reinforcement learning in Isaac Gym with Kalman-filter tracking for dynamic grasping and failure recovery through object reacquisition and grasp retries.
   • Demonstrated sim-to-real transfer to UR5e, achieving 93.3% grasp success in unobstructed scenes and 80% with obstacles, outperforming EARL in both settings.

Co-Author

1. DAP: Diffusion-based Affordance Prediction for Multi-modality Storage

   Haonan Chang, Kowndinya Boyalakuntla, Yuhan Liu, Xinyu Zhang, Liam Schramm, Abdeslam Boularias

   IROS, 2024

   [Paper] [Code]
   • Collected data and evaluated diffusion-based 6D object placement for storage tasks.
   • Achieved 98% book-shelving and 94% can-stacking success in simulation, outperforming the RPDiff placement baseline; demonstrated fruit storage on a KUKA IIWA14.

2. LGMCTS: Language-Guided Monte-Carlo Tree Search for Executable Semantic Object Rearrangement

   Haonan Chang, Kai Gao, Kowndinya Boyalakuntla, Alex Lee, Baichuan Huang, Harish Udhaya Kumar, Jingjin Yu, Abdeslam Boularias

   IROS, 2024

   [Paper] [Project] [Code]
   • Co-developed ELGR-Bench and implemented language parsing, object selection, and evaluation for a planner that combines language guidance with Monte Carlo tree search.
   • Achieved 79.2% executable-plan success on ELGR-Bench and demonstrated language-guided rearrangement on UR5e.

3. Context-Aware Entity Grounding with Open-Vocabulary 3D Scene Graphs

   Haonan Chang, Kowndinya Boyalakuntla, Shiyang Lu, Siwei Cai, Eric Pu Jing, Shreesh Keskar, Shijie Geng, Adeeb Abbas, Lifeng Zhou, Kostas Bekris, Abdeslam Boularias

   CoRL, 2023

   [Paper] [Project] [Code]
   • Built the DOVE-G benchmark and evaluation pipeline for grounding language in open-vocabulary 3D scene graphs, including scene reconstruction and ConceptFusion integration.
   • Achieved 65.8% top-1 3D grounding success on DOVE-G, compared with 41.5% for OVIR-3D.

Teaching experience

Peer review

• IEEE International Conference on Robotics and Automation (ICRA) (2026)
• IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) (2025, 2026)
• Conference on Robot Learning (CoRL) (2026)
• The International Journal of Robotics Research (IJRR) (2026)
• International Conference on Automated Planning and Scheduling (ICAPS) (2026)

Honors & awards

Skills

Real hardware: UR5e, KUKA IIWA14

Simulation platforms: NVIDIA Isaac Gym, Isaac Lab, MuJoCo, PyBullet

Programming & tools: Python, PyTorch, ROS, ROS 2, NumPy, OpenCV, Slurm, Git

Leadership & mentoring

Professional service & activities

2021: Developed an Android application for IIT Tirupati’s convocation.

2020: NSS, IIT Tirupati: 80 service hours; website development and maintenance team lead.