This work is concerned with the development of control policies in order to increase the robustness of robot control systems in the presence of sensor fault or failure. The policies are based on the integration of low sampling rate Heteroceptive Sensor feedback data (i.e., sensor data providing contextual and environmental feedback) in the robot system position servo-lbop where Proprioceptive Sensor data (i.e., sensor data providing feedback about the internal state of the robotic device) is already used to close the feedback loop. Increased robustness is achieved through the dynamic reconfiguration of the sensing subsystem on fault occurrence.

The smooth transition between the different sensing subsystem configuration is performed using Heuristic-based Geometric Redundant Fusion (HGRF), which generates a fused feedback signal based on the relative uncertainty of each feedback signal with respect to the overall sensing uncertainty. The uncertainty of each feedback signal is quantified though the use of an Uncertainty Quantification Metric, which quantifies the signal uncertainty in terms its discrepancy with the deterministic set-point signal of the robotic device.

The crux of the policies lies in defining a coherent and uniform way to optimize the use of the low-rate heteroceptive signal in the high-rate servo-loop. Two policies are proposed in order to compensate for the low sampling rate of the heteroceptive sensor and generate refereed feedback values for all controller signal update instant. The first policy uses Kalman filter-based predictors in order to both time-correlate the slow heteroceptive data, and propagate its information through the following controller update instants. The second policy uses an open-loop polynomial prediction scheme in order to generate heteroceptive measurements for all controller update:​ instants, without time-correlating the lagging heteroceptive signal.

Stability analysis, uncertainty quantification metric performance evaluation and simulation results are presented for both policies. Experimental results are provided for the polynomial policy, based on the implementation of the multirate control policy on a Cartesian manipulator.