Training a simple manipulation policy in sim with RL, and the trials and tribulations of bringing up a new environment

Training an end-to-end locomotion policy that can use depth sensor input, with RL

What's the best way to cut a straight line to fit a curve?

Reimplementing the VINN paper, a non-parametric way of behaviour cloning direct from data.

I created a basic imitation learning policy from the ground up to see what it could learn. Spoiler - not much.

How to start with a robot arm - from first motions in simulation to the real world. Inverse kinematics, trajectories, position control, velocity control, in MuJoCo and on an Ufactory Lite 6.