robosuite

Fine-grained robot control with Jev, physics previews, and configurable LIBERO tasks.

Official Implementation for the paper "SR-AIF: Solving Sparse-Reward Robotic Tasks from Pixels with Active Inference and World Models"

Official Implementation of NeurIPS'23 Paper "Cross-Episodic Curriculum for Transformer Agents"

Agent + reusable skill library for LIBERO manipulation tasks (no RL / VLA / world model; no LLM calls at runtime)

Softbody Mainpulation for Robot based on DRL.

Very minimal Implementation of Pi-Zero (https://www.pi.website/download/pi0.pdf). Vision Action Flow matching model

Analysis of rethink robotics sawyer. Along with reinforcement learning

Autonomous Robotic Arm Control (Franka Panda) using Twin Delayed DDPG (TD3) in Robosuite/MuJoCo. An implementation of Deep Reinforcement Learning for continuous control tasks like Door Opening.

MuJoCo and robosuite platform for multi-arm teleoperation, demonstration collection, replay, and VLA dataset preparation

This work builds on top of mujoco-powered robosuite framework, and is designed for learning object-agnostic control policy for pick-and-place operations on a novel 6-DoF parallel manipulator using deep reinforcement learning. This pipeline utilizes DDPG combined with HER algorithm for training a robust manipulation policy.

Behavioral Cloning and Diffusion Policy for Panda PickPlaceCan manipulation using robosuite, robomimic, RGB observations, and expert demonstrations.

robomimic's imitation learning policies reimplemented in PyTorch as a LeRobot plugin: BC, BC-RNN, BC-Transformer, and BC-VAE, with CLIP language conditioning

The repository for the project assignment "Solving the Pick-and-Place Environment in Robosuite"

Operator-composition vs seed variance in robomimic Multi-Human benchmarks. Preregistered, leave-one-operator-out, matched random control

Agentic LLM planner for robotic manipulation in Robosuite. The model discovers scene state through perception primitives instead of receiving injected object positions, benchmarked against a scripted planner across 300 trials.

Intervention-guided recovery benchmark for simulated robot manipulation

Simulation rollout and evaluation tools for STIR policies in MuJoCo and robosuite.

This repository contains the implementation of a Physically Informed Reinforcement Learning agent. The project utilizes Proximal Policy Optimization (PPO) to train a robotic agent for the "Lift" manipulation task within the Robosuite simulation environment.

Panda manipulation benchmark: RGB-D camera fusion, compact diffusion and ACT-style policies, matched ablations and reproducible evaluation.

WebXR teleoperation and demonstration collection for STIR in MuJoCo and robosuite.