Reliable CPS Design for Unreliable Hardware Platforms

Abstract

Today, many battery-operated cyber-physical systems (CPS) ranging from domestic robots, to drones, and electric vehicles are highly software-intensive. The software in such systems involves multiple feedback control loops that implement different functionality. How these control loops are designed is closely related to both the semiconductor aging of the processors on which the software is run and also the aging of the batteries in these systems. For example, sudden acceleration in an electric vehicle can negatively impact the health of the vehicle’s battery. On the other hand, processors age over time and stress, impacting the execution of control algorithms and thus the control performance. With increasing semiconductor scaling, and our increasing reliance on battery-operated devices, these aging effects are of concern for the lifetime of these devices. Traditionally, the design of the control loops focused only on control-theoretic metrics, related to stability and performance (such as peak overshoot or settling time). In this chapter we show that such controller design techniques that are oblivious of the characteristics of the hardware implementation platform dramatically worsen the battery behaviour and violate the safety requirement with processor aging. However, with proper controller design these effects can be mitigated—thereby improving the lifetime of the devices.