Building Self-Adaptive Cyber Physical Systems Using Unreliable Components

1. Introduction Modern automobiles, featuring increased convenience, safety, and energy savings, have become more and more reliant on the outcomes of the IT industry. Existing services such as vehicle tracking, navigation and emergency notification envision a promising future of more powerful computing-assisted automobile systems. Some examples include remote vehicle controllers, short range braking systems, and eventually, the self-guided automobile could drive instead of human beings. However, with established techniques, designers could barely implement more advanced systems due to several specific requirements of future CPS. First, the system has to be very reliable. However, the sensing data are not reliable given the harsh physical world where sensors are deployed and operated. Second, adaptation is the key to the success of CPS, but to the best of our knowledge, we haven't seen a mature theoretical model that can guide the adaption. Finally, as technologies have gradually immersed into our personal lives, how much should we trust and rely on the technology, and who will be responsible for the mistakes?