Robot Control The job of the robot controller is to convert the task specification to forces and torques at the actuators. Control strategies include motion control, force control, hybrid motion/force control, or impedance control. Which of these behaviors is appropriate depends on both the task and the environment. For carrying... [Read More]

Why using OpenAI Spinning Up? Reinforcement learning results are tricky to reproduce: performance is very noisy, algorithms have many moving parts which allow for subtle bugs, and many papers don’t report all the required tricks. With code bases like OpenAI Baselines or OpenAI Spinning Up, researchers can spend less time... [Read More]

Why creating an environment for Gym? OpenAI Gym is the de facto toolkit for reinforcement learning research. Researchers use Gym to compare their algorithms for its growing collection of benchmark problems that expose a common interface. Two important design decisions have been made for this common interface: Two core concepts... [Read More]

Many of the standard environments for evaluating continuous control reinforcement learning algorithms are built using the MuJoCo physics engine, a paid and licensed software. Bullet Physics provides a free and open source alternative to physics simulation. [Read More]

I have recently started using Vim after using Emacs for a while. I have Linux machine and a mac laptop at home and a Windows machine at work. I code mostly in Python and C++ and I got frustrated when every time when I switch my machine I have to... [Read More]