Cyber-physical systems (CPS) are widely used for safety-critical applications, including automotive electronics and medical equipment, where reliability and performance are important issues to consider. In the meantime, as the feature sizes of VLSI circuits used in CPS continue to shrink, the power density has increased rapidly, causing a significant rise in chip temperatures. These elevated temperatures have an adverse impact on the system's reliability, performance, and power efficiency. Therefore, it is essential to consider thermal effects during the CPS design process, in particular at the system design level. This paper discusses the challenges of designing and optimizing a CPS, with a focus on the interplay between thermal issues and the stringent reliability and real-time requirements demanded by safety-critical CPS applications. Through this discussion, we will explore the various issues related to CPS design and optimization in the context of different thermal problems. We will also present a temperature-aware task scheduling and mapping technique for heterogeneous CPS platforms consisting of microprocessor cores, GPU processors, and FPGA units. By carefully analyzing and considering thermal impacts during the design process, we can ensure that the CPS to be designed meets the high standards required for safety-critical applications.