A Robust Control Scheme for Autonomous Vehicles under Sensor Perception Bias

Abstract

The challenge of ensuring robust autonomous driving control under sensor perception bias arises from the coupling among uncertainties in perception, hydraulics, computation, and actuation. A single deterministic controller or classifier cannot guarantee safety in such conditions. This paper proposes a diagnostic and control framework that integrates bias-aware residual construction, model-predictive reasoning, and conservative fallback rules using multi-sensor inputs. Parametrisation will be employed to obtain the study data instead of the field claim data, and all indicators are design variables for repeated engineering evaluations. The five tables are sensor channel information, disturbance examples, diagnostic thresholds, control responses and robustness indicators; the two formulas are the residual score and the weighted stability index. Based on the above results, the proposed framework can reduce false alarms and ensure the normal operation of the system in all circumstances; that is to say, if the input signal is biased, there is a time lag, or some data are missing. Finally, the added content will provide the basis for the fault, model limitation and practical control measures to form an auditable process in accordance with the requirements of regular engineering publications.