• aditya mehrotra.

chip updates: we're doing this wrong [updates]

So here's my thinking. We're really over complicating our lives by trying to develop everything about CHIP completely form scratch. And I know we keep flip-flopping positions but really it's true, we need a better plan. If we want CHIP to work. What we need to do is have the IMU stabilization active all the time right, that's the end-goal. So we need to start with the IMU stabilization.

From tomorrow, forget test mode, forget everything else, we're going to do simply one thing. We're going to work on the IMU node and just have two legs on the ground (any two diagonal legs) and just have ONE FOCUS, make the robot try to balance on these two legs. THAT's IT. Using PID or whatever techniques come to mind, ONLY focus on having it balance on two legs. No standing no sitting no etc. We'll retract the other two legs and hold them completely stationary. No walking no gates no etc.

JUST MAKE IT BALANCE ON TWO LEGS because we need to START somewhere and then do incremental development you know. Since the legs can move very fast now we might actually be able to have the robot balance itself on two legs by jerking those two legs forwards/backwards and side/side.

This is another interesting thing I found: https://www.mathworks.com/matlabcentral/fileexchange/64237-running-robot-model-in-simscape

It's a Matlab four-legged robot simulation. I don't know what it does, I don't know if we can even use it, but it could be worth a shot if what we're trying isn't working.

We want to try STRAIGHT PID first and see can we just get this thing to balance on two legs that's the ONLY IMPORTANT QUESTION right now.

Why I keep repeating that? We keep falling into this trap of trying to do way too much at once. So all we're going to do is when we set the robot into test mode, it will raise two feet and the robot will begin balancing on the other two legs (in theory). And it'll all be hardcoded which legs lift and which legs balance. The focus on on the BALANCE right now that's it (again repeating so we don't forget).

WHY? Because the moment we nail this, we will have solved CHIP for the most part. If we can balance on two legs the other two can move however we want them to or if three are on the ground the balancing will be good enough so that the robot will never fall no matter how many legs we pick up you know. THATS THE GOAL. Develop enough of a sense of balance first that no matter how we move the legs the robot will stay upright.

Here's good video on how spot works - we're going to watch it to get some idea of how it works as is what this video is about!

Okay I don't think this'll teach us how spot walks because it's some seriously advanced stuff but it's cool anyways. BUT YES THATS THE PLAN. BALANCE ON TWO LEGS FIRST. There's a very interesting part around ~8min in this video which describes how these robots walk by FALLING. I don't know how to do that. It might not be possible to get the robot to balance on two legs. But we shall try.

See ~9min for a nice dance mode. See ~12min for use of April tags and etc.

So yes, balance on two legs, that's the plan, yeet. That's the ONLY GOAL RIGHT NOW LETS NOT GET SIDE TRACKED.

#updates #omgrobots #plans #re_plan

0 views0 comments
© copyright 2019 | aditya mehrotra