Task Scheduling for Real-Time Object Detection: Methods and Performance Comparison in ADAS Applications

Computer Vision has changed how a lot of smart systems function. Cyber Physical systems, such as Advanced Driver Assistance Systems (ADAS), make use of this technology to see their surroundings and act accordingly, but since they are computer systems, they need to process information and do one action at a time. Deciding which task to process and respond to requires implementing a scheduler that can prioritize the tasks in a manner we deem appropriate. We must also acknowledge that some tasks have deadlines that need to be met to ensure safety. Given all this, Real-time scheduling proves to be an important subject in this area. In this paper, we will introduce and compare different scheduling algorithms and see how they perform under machine vision tasks, and we also propose a deadline handling technique to reduce further deadline misses.