Benchmark

In this section, we bring detailed explanations for the 3 AI-enabled CPS in our experiment.

1. Adaptive Cruise Control (ACC)

The system is released by Mathworks. This system targets to control the acceleration of the ego vehicle to keep the relative distance Drel to the lead car greater than a safety distance Dsafe. The safety distance is dynamically changed based on the relative velocity between two cars. In addition, when the safety distance is assured, the ego car should approach to use-set cruising velocity vtarget.

2. Abstract Fuel Control (AFC)

AFC is released by Toyota which simulates a powertrain system of a vehicle. The external inputs are pedal angle and engine speed, and the fuel injection rate to a fuel cylinder is controlled by the system. The air-fuel-ratio AF inside the cylinder is required to maintain a reference value AFref to achieve complete gasoline fuel combustion. The optimized stoichiometric air-fuel-ratio is about 14.7 and the deviation should be limited within 0.1.

3. Exothermic Chemical Reactor (CSTR)

CSTR is originally released by Mathworks which is a widely used system in various chemical industry domains. The concentration of the reagent in the exit stream of the reactor is expected to sustain the setpoint. When the reactor is triggered from a low transformation ratio to a high transformation ratio, the concentration setpoint will change accordingly. In addition, the error of the concentration should be less than 0.35 during the entire transformation.