PyBullet Robotics Simulator

PyBullet is a Python module for physics simulation, robotics, and machine learning reinforcement learning, built on top of the Bullet Physics SDK (Bullet 3). Originally developed by Erwin Coumas at Google, PyBullet provides a Pythonic interface to the Bullet physics engine for robotics simulation and reinforcement learning research. Key features: 3D rigid and soft body dynamics simulation with Bullet Physics Engine supporting collision detection, constraint solving, and articulated body dynamics. Robot simulation with URDF, SDF, and MJCF file format loading for importing robot models from ROS, Gazebo, and MuJoCo. Inverse kinematics via IK solver with null-space control and joint limit handling for robotic arm motion planning. Inverse dynamics computing required joint torques for desired trajectories. Ray casting and batch ray casting for sensor simulation (lidar, depth, contact). Camera rendering with software-based OpenGL renderer for vision-based RL tasks. Soft body and cloth simulation using finite element method FEM and position-based dynamics PBD. Locomotion benchmarks including DeepMimic, Minitaur quadruped, PyBullet humanoid, Laikago, and ANYmal robot simulations. Integration with TensorFlow, PyTorch, and Stable Baselines for reinforcement learning training pipelines. Real-time simulation with configurable time stepping and fixed timestep control. VR support via OpenVR for teleoperation and immersive simulation environments. Cross-platform on Windows, Linux, and macOS. Open source under zlib license.