Bikram Pandit

I am a PhD candidate in Artificial Intelligence at Oregon State University, in the Dynamic Robotics and AI Lab, advised by Alan Fern. My research focuses on reinforcement learning for legged, wheeled, and hybrid single- and multi-robot systems, with a proven track record of transferring policies from simulation to hardware, across Cassie, D1, Spot, Unitree Go2, Digit, and Unitree H1.

I previously interned at the RAI Institute (formerly Boston Dynamics AI Institute), working on humanoid locomotion and loco-manipulation, and at MiCROTEC as a Machine Learning intern. Before my PhD, I spent several years as a Senior Software Engineer at Digital Product Labs, designing and deploying production ML systems.

news

Aug 20, 2026	📝 Filed a patent, “Systems and Methods for Decentralized Multi-Robot Cooperative Transport” (under review).
Mar 24, 2026	🏆 Our team won First Place in the DARPA TIAMAT Challenge, with my contribution on robust Spot policies for high-speed exploration, fall recovery, and obstacle avoidance.
Jan 31, 2026	📄 New paper, decPLM, on decentralized cooperative transport with teams of quadruped robots, accepted to ICRA 2026.
Aug 08, 2025	🎤 Our paper on non-conflicting energy minimization for RL-based robot control was accepted as an Oral at CoRL 2025.
Sep 23, 2024	🦾 Started a research internship at the RAI Institute (formerly Boston Dynamics AI Institute), working on humanoid locomotion and loco-manipulation.
Sep 05, 2024	✨ “Learning Decentralized Multi-Biped Control for Payload Transport” was accepted as a Spotlight at CoRL 2024.
Apr 11, 2024	🎙️ Won 1st Prize for the lightning talk “Learning Vision-Based Bipedal Locomotion for Challenging Terrain” at the OSU AI Meetup for Industry and Alumni.
Jan 29, 2024	🏅 “Learning Vision-Based Bipedal Locomotion for Challenging Terrain” was a Best Paper Finalist at ICRA 2024.
Jan 30, 2023	🥈 Our team own 2nd Prize in a hackathon organized by the AgAID Institute (sponsored by Microsoft), competing against 32 teams from 8+ universities.

selected publications

ICRA
Multi-Quadruped Cooperative Object Transport: Learning Decentralized Pinch-Lift-Move

Bikram Pandit, Aayam Kumar Shrestha, and Alan Fern

International Conference on Robotics and Automation (ICRA), 2026

Abs arXiv Bib Code Website

We study decentralized cooperative transport using teams of N quadruped robots with arms that must pinch, lift, and move ungraspable objects through physical contact alone. Unlike prior work that relies on rigid mechanical coupling between robots and objects, we address the more challenging setting where mechanically independent robots must coordinate through contact forces alone without any communication or centralized control. Our approach enables coordination through shared policy parameters and implicit synchronization cues, scaling to arbitrary team sizes without retraining.
@inproceedings{pandit2026decplm, title = {Multi-Quadruped Cooperative Object Transport: Learning Decentralized Pinch-Lift-Move}, author = {Pandit, Bikram and Shrestha, Aayam Kumar and Fern, Alan}, booktitle = {International Conference on Robotics and Automation (ICRA)}, year = {2026}, }
CoRL
Oral
Non-conflicting Energy Minimization in Reinforcement Learning based Robot Control

Skand Peri, Akhil Perincherry^*, Bikram Pandit^*, and Stefan Lee

Conference on Robot Learning (CoRL), 2025

Abs arXiv Bib Code Website

Efficient robot control often requires balancing task performance with energy expenditure. We propose a hyperparameter-free gradient optimization method to minimize energy expenditure without conflicting with task performance, applying policy gradient projection between task and energy objectives. We demonstrate a reduction of 64% energy usage while maintaining comparable task performance, with sim-to-real validation on a Unitree Go2 quadruped.
@inproceedings{peri2025pegrad, abbr2 = {Oral}, title = {Non-conflicting Energy Minimization in Reinforcement Learning based Robot Control}, author = {Peri, Skand and Perincherry, Akhil and Pandit, Bikram and Lee, Stefan}, booktitle = {Conference on Robot Learning (CoRL)}, year = {2025}, }
CoRL
Spotlight
Learning Decentralized Multi-Biped Control for Payload Transport

Bikram Pandit, Ashutosh Gupta, Mohitvishnu S. Gadde, Addison Johnson, Aayam Kumar Shrestha, Helei Duan, Jeremy Dao, and Alan Fern

Conference on Robot Learning (CoRL), 2024

Abs arXiv Bib Code Website

We replace wheels with multiple bipedal robots attached to a carrier for payload transport on rough terrain. Our main contribution is a decentralized controller that works with varying numbers and configurations of robots without retraining, trained via reinforcement learning in simulation and transferred to the real world, with results shown on systems of two and three Cassie robots.
@inproceedings{pandit2025decmbc, abbr2 = {Spotlight}, title = {Learning Decentralized Multi-Biped Control for Payload Transport}, author = {Pandit, Bikram and Gupta, Ashutosh and Gadde, Mohitvishnu S. and Johnson, Addison and Shrestha, Aayam Kumar and Duan, Helei and Dao, Jeremy and Fern, Alan}, booktitle = {Conference on Robot Learning (CoRL)}, year = {2024}, }
ICRA
Best Paper Finalist
Learning Vision-Based Bipedal Locomotion for Challenging Terrain

Helei Duan, Bikram Pandit, Mohitvishnu S. Gadde, Bart van Marum, Jeremy Dao, Chanho Kim, and Alan Fern

International Conference on Robotics and Automation (ICRA), 2024

Abs arXiv Bib

Reinforcement learning for bipedal locomotion has demonstrated robust gaits over moderate terrains using proprioceptive sensing, but blind controllers fail in environments requiring visual perception. We propose a fully-learned system that enables bipedal robots to react to local terrain while maintaining travel speed and direction, achieving successful sim-to-real transfer without explicit pose estimation.
@inproceedings{duan2024visionbipedal, abbr2 = {Best Paper Finalist}, title = {Learning Vision-Based Bipedal Locomotion for Challenging Terrain}, author = {Duan, Helei and Pandit, Bikram and Gadde, Mohitvishnu S. and {van Marum}, Bart and Dao, Jeremy and Kim, Chanho and Fern, Alan}, booktitle = {International Conference on Robotics and Automation (ICRA)}, year = {2024}, }
CoRL
Simple Masked Training Strategies Yield Control Policies That Are Robust to Sensor Failure

Skand Peri, Bikram Pandit, Chanho Kim, Fuxin Li, and Stefan Lee

Conference on Robot Learning (CoRL), 2024

Abs Bib HTML

Sensor failures are common in real-world robot deployments, leading to potential catastrophic outcomes. We propose training reinforcement learning policies that are robust to sensor failures using a multimodal encoder and a training strategy that simulates sensor failures by randomly dropping sensor modalities, evaluated on bipedal locomotion and robotic manipulation.
@inproceedings{peri2024maskedtraining, title = {Simple Masked Training Strategies Yield Control Policies That Are Robust to Sensor Failure}, author = {Peri, Skand and Pandit, Bikram and Kim, Chanho and Li, Fuxin and Lee, Stefan}, booktitle = {Conference on Robot Learning (CoRL)}, year = {2024}, }