I am currently a lead software engineer for an SME with over 10 years of experience building mainly Web apps and services (backend, frontend, data engineering ... etc). I have grown an interest in hardware (I have a CS degree) and ever since then I have been trying to self-learn robotics.
How can I self-learn robotics in a way that allows me to switch careers? I'd like to get into the field and move away from Web dev.
I would say that on one end, you have high level algorithm related to motion planning, object detection, navigation, SLAM, autonomous driving, kinematics and inverse kinematics, and on the other end, so to speak, the low level implementation of how to drive a motor, how to generate a PWM from a microcontroller, and it is pretty much embedded development.
Probably the first step would be to decide on which part you want to focus on primarily, while having a good view on the whole stack help you gives you some credibility during an interview.
For the high end, there is a number of MOOC, and you can play around with some like ROS (or ROS2), for the embedded part, you can probably look into some some microcontrollers and motion control.
You can also probably look for some job where you could reuse your existing skill set, like a web interface to pilot a fleet of robots.