Constructing dependable certainty grids from unreliable sensor data

Measurements from sensors as they are used for robotic grid map applications typically show behavior like degradation or discalibration over time, which affects the quality of the generated maps. This paper presents two novel algorithms for the generation of certainty grids dealing with this behavior. The first algorithm named Fault-Tolerant Certainty Grid (FTCG) performs voting over multiple sensor readings. This approach removes up to (n-1)/2 faulty measurements for grid cells that are updated by n independent sensors, however requires that each grid cell is covered by at least three different independent sensors. The second algorithm named Robust Certainty Grid (RCG) uses a sensor validation method that detects abnormal sensor measurements and adjusts a confidence value for each sensor. This method supports also reintegration of recovered sensors from transient faults and sensor maintenance by providing a measurement for the operability of a sensor. The RCG algorithm works with at least three sensors with a partially overlapping sensing range and needs fewer sensor inputs and less memory than the FTCG approach. Results from simulation and an experimental evaluation on an autonomous mobile robot show that under the presence of unreliable sensor data, both algorithms perform better than the Bayesian approach typically used for certainty grids.