Good looking flock of robots ya got there :) A few observations:
1) Actuators entirely enclosed now? That affect stator/controller temps? 2) Knee pulleys casted out of resin or fiberglass now? 3) Black shoulder plates: my first thought was collision pads, but I could be convinced that the back right plate on the leftmost robot at 0:38 had a white wire coming out from it. Any function beyond protection? 4) That's quite a few radios in a tiny area. Any issues with communication interference?
1) The actuators are enclosed. Thermals had nothing to do with why the original ones were open, that was entirely for weight reduction. They are enclosed now because these robots are going to other research groups and I don't want to have to worry about crap getting in the gearboxes. I'm sure it makes the motors a little bit hotter, but it hasn't been enough to cause problems.
2) No, the knee pulleys are still post-machined misumi GT2 pulleys. There are UHMW belt protectors that snap onto the knee around them to protect the belt against knee-ground collisions. I've had a set of those on the original mini cheetah for the last year + and they've been super durable.
3) They're just shoulder protectors. They have a layer of hard nylon on the outside with an inner layer of poron foam to spread out impacts
4) Didn't notice any radio problems. They're all Frsky receivers/transmitters, they've been really solid.
I've been following your progress for some time now and I wanted to thank you for sharing your content. It turns out that https://www.youtube.com/watch?v=LQD9JmwfG2A
Great work as always! I just published the first motion test video of a quadruped I built recently. It's a bit bigger than the Mini Cheetah. You can find it here: https://www.instagram.com/778labs/
Would love to hear your thoughts.
Also - are there any publications as to the control algorithms on the Mini Cheetah? I have read somewhere that you are using some MPC approach, but haven't seen any specifics anywhere. Is everyone solving this problem independently, or is there some material out there I can learn from?
Good looking flock of robots ya got there :) A few observations:
ReplyDelete1) Actuators entirely enclosed now? That affect stator/controller temps?
2) Knee pulleys casted out of resin or fiberglass now?
3) Black shoulder plates: my first thought was collision pads, but I could be convinced that the back right plate on the leftmost robot at 0:38 had a white wire coming out from it. Any function beyond protection?
4) That's quite a few radios in a tiny area. Any issues with communication interference?
Love it! Thanks for sharing
1) The actuators are enclosed. Thermals had nothing to do with why the original ones were open, that was entirely for weight reduction. They are enclosed now because these robots are going to other research groups and I don't want to have to worry about crap getting in the gearboxes. I'm sure it makes the motors a little bit hotter, but it hasn't been enough to cause problems.
Delete2) No, the knee pulleys are still post-machined misumi GT2 pulleys. There are UHMW belt protectors that snap onto the knee around them to protect the belt against knee-ground collisions. I've had a set of those on the original mini cheetah for the last year + and they've been super durable.
3) They're just shoulder protectors. They have a layer of hard nylon on the outside with an inner layer of poron foam to spread out impacts
4) Didn't notice any radio problems. They're all Frsky receivers/transmitters, they've been really solid.
Hi Ben,
ReplyDeleteI've been following your progress for some time now and I wanted to thank you for sharing your content. It turns out that https://www.youtube.com/watch?v=LQD9JmwfG2A
oh well ... just wanted to mention
How was Macao?
Hello Ben,
ReplyDeleteGreat work as always! I just published the first motion test video of a quadruped I built recently. It's a bit bigger than the Mini Cheetah. You can find it here: https://www.instagram.com/778labs/
Would love to hear your thoughts.
Also - are there any publications as to the control algorithms on the Mini Cheetah? I have read somewhere that you are using some MPC approach, but haven't seen any specifics anywhere. Is everyone solving this problem independently, or is there some material out there I can learn from?