On the Effects of Measurement Uncertainty in Optimal Control of Contact Interactions
This addresses the challenge of uncertain contact interactions in robotics, such as locomotion and manipulation, by highlighting the distinct effects of measurement uncertainty, though it is incremental as it builds on existing stochastic optimal control methods.
The paper tackled the problem of measurement uncertainty in robotic contact interactions, developing a stochastic optimal control algorithm that incorporates measurement noise and observer dynamics, and found in simulations that measurement uncertainty leads to low impedance behaviors, contrasting with the stiff behaviors from process noise.
Stochastic Optimal Control (SOC) typically considers noise only in the process model, i.e. unknown disturbances. However, in many robotic applications involving interaction with the environment, such as locomotion and manipulation, uncertainty also comes from lack of precise knowledge of the world, which is not an actual disturbance. We analyze the effects of also considering noise in the measurement model, by developing a SOC algorithm based on risk-sensitive control, that includes the dynamics of an observer in such a way that the control law explicitly depends on the current measurement uncertainty. In simulation results on a simple 2D manipulator, we have observed that measurement uncertainty leads to low impedance behaviors, a result in contrast with the effects of process noise that creates stiff behaviors. This suggests that taking into account measurement uncertainty could be a potentially very interesting way to approach problems involving uncertain contact interactions.