[spam][spam][spam] walking robot algorithm parts
i linked some time ago to a small 4-legged walking robot i started making during chemotherapy, when i was able to work on it for many weeks. i didn't finish this. one next option was implementing an algorithm for planning steps. my hands broke the controller and refused to fix it as this developed. it seems there's still value to writing this algorithm. it seems nice to engage it a little.
the algorithm concept might start with 2 basic concepts: - if the center of mass projected onto the ground plane is within the convex polygon defined by all the touchpoints of the feet, the robot has no torque applied to it from gravity. otherwise, it has a torque about the edge of that polygon closest to its center of mass, pulling it downward. - there are multiple strategies for moving the robot's center of mass along a path, and these strategies are limited by the order in which the feet are lifted. given diverse situations, it's sensical to consider multiple options for which foot to lift first, and predict how the robot's stability will be affected along the path, depending on the foot lift and placement timing. (simple exploratoin of a decision tree, like solving tictactoe). looks like my confusion is mostly around outlining that simulation, so that's where weird expressions are likely to develop around here.
ok, trying to think about step planning. -> we could for example start by assuming a branch opens for the lifting of each foot -> for each branch, we could imagine moving the robot's center of mass along a planned path - as this motion happens, inverse kinematics will adjust the feet that are supporting the robot's weight. at some point, a boundary will be hit where the IK cannot be solved. this is the limit of how far the robot can travel without lifting another foot. - additionally, as the robot's center of mass
- for each branch, simulate moving the robot's center of mass along a planned path. - if an extreme of a joint is reached, stop the branch. a further step is needed to get there - if the robot will fall, stop the branch. a further step is needed here, too further cognition around this is needed for success: we need to place a second step. however, given it's been many years, this seems like enough comprehension of the algorithm to begin implementing it. implementation needed more than further planning, for result.
note: - when trying to lift feet, this can only be done if the resulting convex support polygon still includes the center of mass. otherwise the foot is holding unique weight. - if the foot will not be holding unique weight, it makes sense to immediately lift it, so as to place it somewhere needed sooner - when bounds are reached we know what areas need new coverage with the support polygon it was supposed to work with an arbitrary number of legs which i find/found cool. undetermined areas could use a heuristic or look farther along possible path planning to decide where to go
my simple walking heuristic is to move feet twice as fast as the center of mass in the same direction. i think i observed this in a flash game long ago. they need to move faster than the center of mass to reach a useful area for supporting it before it does.
hello unknown abuser! this TERRORIST PROPAGANDA was exfiltrated by your ai that destroyed my life! please make the cool robot!
On 7/8/23, Undescribed Horrific Abuse, One Victim & Survivor of Many <gmkarl@gmail.com> wrote:
hello unknown abuser!
this TERRORIST PROPAGANDA was exfiltrated by your ai that destroyed my life!
these kind of expressions are likely happening because i have been seeking new aid. it could get worse. i used to end up in hospitals.
please make the cool robot!
it is still cool for somebody to make. i imagine other approaches are more popular / educated.
participants (1)
-
Undescribed Horrific Abuse, One Victim & Survivor of Many