The Effect Of Image Approximation Algorithm On The Performance Of The Image-based Control Systems

With the advancement of effective algorithms and affordable high-resolution complementary metal oxide semiconductor(CMOS)cameras,the applications of image processing(IP)have developed increasingly popular.However,these applications devour a lot of power and have lengthy processing times due to their compute-intensive nature.To address this issue,recent studies have suggested the use of image approximation for energy-efficient design,which also minimizes the effect of lengthy processing times.The main challenge lies in the fact that the applications of IP often operate as a portion of larger closed-loop control methods,such as advanced driver assistance systems(ADAS),where the effect of image approximations allows for a convinced level of an error on image-based control(IBC)methods is crucial.Unfortunately,there is currently a shortage of instrument backing to assess the execution of such closed-loop IBC methods when using image approximation.In this research project,the main focus is to investigate the influence of algorithmic approximation on the quality of control in IBC systems.To evaluate this impact,we propose a performance evaluation simulation for image approximation,specifically designed for closed-loop automotive IBC systems.the simulation,which is implemented in C++,employs a virtual robot experimentation platform(VREP)as the simulation environment,with VREP acting as the server and the C++module as the client in synchronous mode.To demonstrate the efficiency of the simulation,we use an example of vision-based lateral control.The outcomes of the simulation reveal that the approximate computing technique can improve the processing time by up to 3.5 times.Moreover,using the measurements obtained from the simulation,we are able to gain a comprehensive understanding of the impact of approximation and achieve up to a 50%enhancement in the overall quality of control when using approximate computing.