ESTIMATION AND CONTROL BY MODELING THE INTERACTIONS BETWEEN PROPRIOCEPTIVE FRAME AND THE REAL WORLD
Séance du mardi 14 novembre 2017, Salle L312, 09h00-10h00.
Mehdi BENALLEGUE, AIST, Tsukuba, Ibakari, Japan
An important part of the robot-state is usually reliably obtained through the internal proprioceptive sensors, typically joint encoders. We can see it as a direct measurement of the kinematics in a "proprioceptive" or ideal frame. Most robots are also usually actuated in this frame. The challenge for estimation and control world reduces then to deal with the deviations between this frame and the world frame. In this talk, we show how the interactions between these frames is a simple set of equations involving external forces and moments. And how this model can allow reconstructing a large set of state variables in a single estimator, consistent with the dynamics and with sensor measurements. We finally present a controller based on this concept allowing for an efficient stabilizer.
Mehdi BENALLEGUE, AIST, Tsukuba, Ibakari, Japan
An important part of the robot-state is usually reliably obtained through the internal proprioceptive sensors, typically joint encoders. We can see it as a direct measurement of the kinematics in a "proprioceptive" or ideal frame. Most robots are also usually actuated in this frame. The challenge for estimation and control world reduces then to deal with the deviations between this frame and the world frame. In this talk, we show how the interactions between these frames is a simple set of equations involving external forces and moments. And how this model can allow reconstructing a large set of state variables in a single estimator, consistent with the dynamics and with sensor measurements. We finally present a controller based on this concept allowing for an efficient stabilizer.