Dynamic Memory Management in Critical Embedded Software

Memory management has always been a delicate issue in critical embedded software because memory is often a scarce resource and many of the typical software errors jeopardizing the integrity of the execution of the software are related to memory mismanagement. Furthermore, critical software has had a tendency to grow in size and complexity in recent years, because it is using more and more complex algorithms in the critical parts of a system. The push towards autonomous mobility is a good example of the drivers for complexity reaching the most critical parts of software controlling such systems. This added complexity requires added flexibility in memory management that is not compatible with the traditional memory management techniques used for critical embedded software. In this paper we will first go over the traditional memory management limitations and the reasons behind them, we will then explore possibilities for going beyond them while being able to provide a high level of guarantees of correctness with regard to memory usage.