Approach and Algorithm for Evaluating the Allowed Signal/Noise Ratio of Robotic Lidars under External Influences

Abstract

Purpose or research. The aim of the study is to ensure the safe operation of robotics by developing methods, approaches and algorithms for information processing, and describing their functioning.Methods. The paper proposes an approach to estimation allowed signal/noise ratio (SNR) for robotic LiDARs based on the predetermined probability of occurrence of «false alarm» under unintended influences. The synthesized probabilistic approach is based on the physical fundaments of infrared radiation, and the Bayesian theory using the Neyman-Pearson criterion. The feature of the proposed approach is the use of the given threshold of «false alarm» occurrence, and the probability of occurrence of interference in the analytical apparatus, as well as consideration of the characteristics of photodetectors. This allows expressing analytically and calculating the value of the allowed SNR when stabilizing the level of «false alarms» against background noise caused by this type of interference.Results. The formed and presented dependencies can be used as one of the operating characteristics in the development and selection of optoelectronic system of LiDAR’s measurement system. Based on the fixed value of «false alarm», and the resulting graphical expression of the operating characteristic (obtained characteristics) it is possible to choose a LiDARs system with necessary technical parameters.Conclusion. The probabilistic approach and the corresponding algorithm for selecting the threshold SNR value based on the Neyman-Pearson criterion were developed. The approach allows minimizing the probability of «ignoring» the object when scanning, since the probability of «false alarm» does not exceed the given threshold value. Mathematical and methodological support for the design of LiDARs is presented, taking into account a priori estimation of the allowed SNR value, and the probability of reflected pulse detection, without preliminary estimates of probabilistic characteristics of object detection. The presented algorithm has a set of raw data (in the form of the values of the received signal with a noise component) as an input. Its output is represented by a set of error probability dependencies for different SNR thresholds.