Search Results for "tasks"

Microduck RL provides reinforcement-learning environments for training behaviors for Pollen Robotics' small bipedal Microduck robot. It is built on mjlab with MuJoCo Warp and trains PPO policies at a 50 Hz control rate. The repository includes tasks for walking, standing up, sitting, kicking, rolling, skating, crouching, and other physical behaviors. Its sim-to-real setup models actuator physics, backlash, domain randomization, and reward design. Trained policies can be exported to ONNX and deployed through the separate Microduck runtime. Training normally uses CUDA hardware, while supported workflows can offload jobs to Hugging Face Jobs when a local GPU is unavailable.

NVIDIA Isaac‑GR00T N1.5 is an open-source foundation model engineered for generalized humanoid robot reasoning and manipulation skills. It accepts multimodal inputs—such as language and images—and uses a diffusion transformer architecture built upon vision-language encoders, enabling adaptive robot behaviors across diverse environments. It is designed to be customizable via post-training with real or synthetic data. The vision-language model remains frozen during both pretraining and...

...Its primary goal is to make all available open-source robotic data interoperable by representing them using the RLDS (Reinforcement Learning Dataset Structure) episode format. This enables seamless integration for training, evaluation, and model development across diverse robotic tasks and embodiments. The dataset aggregates contributions from multiple open-source robotic projects, all harmonized under a single unified data schema. The repository also provides Colab notebooks for dataset visualization, batching, and model inference, along with pretrained model checkpoints such as RT-1-X, a multitask robotic transformer model trained on this data.

Roboschool is a set of open source robot simulation environments for reinforcement learning, created as an alternative to the Mujoco physics engine. It integrates with OpenAI Gym and provides a variety of continuous control tasks, including humanoid locomotion, quadrupeds, and robotic arms. The library is built on the Bullet Physics engine, making it accessible without the licensing requirements of Mujoco. Roboschool includes training scripts and examples for applying reinforcement learning algorithms to its environments. While the project has since been deprecated in favor of more modern frameworks, it remains historically significant as a bridge between early reinforcement learning research and scalable, open-access environments. ...

PyRobot is a high-level robotics API that simplifies controlling real and simulated robots through a consistent Python interface. Instead of fighting ROS internals, you call clear methods for things like base navigation, arm manipulation, and camera capture, and PyRobot handles the plumbing. It was built to make research repeatable on accessible platforms—most notably low-cost mobile manipulators—while still scaling to richer hardware. The library exposes canonical abstractions (base, arm,...

To design a complete, wheeled robot, capable of accomplishing simple tasks, manipulate objects and have human interaction capabilities. Includes electronic board design, firmware and AI programming, computer-robot interaction software and mechanics.

...It includes a nifty stopwatch widget as a metaphor for measuring performance. The monitoring is done via event messages which are displayed in real time in a GUI dialog and stored in a JavaDB database. Any number of tasks (Runnables) can be launched and monitored simultaneously. The user / programmer of the API Jar designs their own messages, specifying the set of properties and their types which will be sent with each event they want to report. The properties in the custom message are the fields in a log table created specifically for that task. ...

...KIPR, KISS Institute for Practical Robotics is an organization that facilitates the learning of robotics programming in middle and high school students to peak an interest in computer science. To do this they hold competitions around the world called botball where students program autonomous land robots to perform various sets of tasks. KIPR, now wants to expand and leave the ground by integrating flying robots (the AR Drone) into the tasks for added interest, fun, and challenge for students.

A simulation environment for 2D robots to interact and perform tasks in. Meant to be educational and demonstrate emergent behaviors. No download necessary! See demos on project website!

This project consists in a simulation software of robot A.I. It is aimed at comparing the efficiency of robot intelligence against movement tasks between fixed checkpoints in a logical world.