Intelligent robots are widely used in cyber-physical systems (CPS). In this CPS, intelligent robots with different performance levels act as actuators to achieve tasks with different difficulty levels. On one hand, the robot with the good performance can efficiently achieve a given control task and obtain the high utility. However, the high cost of this task by the robot might be generated as well. On the other hand, during task execution, the robot needs to maintain the reliable communication with the control center in order to receive the precise instructions in time. Therefore, the objective of our optimization in this paper is to maximize the total reward of task execution by balancing utility and cost in the whole CPS. An optimal scheme of jointly allocating the tasks and communication resources is proposed. The scheme obtains the power allocation and the task allocation based on convex optimization theory. Numerical results indicate that the proposed joint allocation scheme can significantly improve the total reward compared to other schemes.