SAFETY MECHANISMS FOR ARTIFICIAL INTELLIGENCE UNITS USED IN SAFETY CRITICAL APPLICATIONS

A method for safety monitoring of an artificial intelligence, Al, processing unit comprising: providing a safety unit for sending a first signal to the Al processing unit S1; determining, by the Al processing unit, a second signal in response to the first signal; transmitting, by the Al processing unit to the safety unit the second signal and a heartbeat signal of the Al processing unit S2, the heartbeat signal representing a time sequence of the first and second signals; determining whether the second signal is correct S3 and whether the time sequence S4 is correct. Preferably, if either the time sequence of second signal are incorrect, the AI may be shut down or reset S5. Preferably, a time period between a falling edge and a rising edge of the heartbeat signal may be used to verify that the clock of the AI processing unit is correct.