The EU-funded project RADIO brings forward a new healthcare paradigm according to which a mobile robotic platform can act as an assistant to an elderly person in his/her domestic environment. The main goal of the robot is to detect ADL (Activities of Daily Life) related to basic self-care tasks, such as sleeping and taking medications, as well as instrumental ADL related to housework. ADL detection is based on visual, depth, and audio signal analysis as well as their fusion. However, robot assistance in everyday living suffers from limited autonomy dictated by the robot battery, e.g., the robot has to constantly know where the person is and to be able to move if the person moves to a new location or another room. In this chapter, we present the line of research followed in the RADIO project in order to reduce the usage of the power-hungry processing components and, consequently, the need for revisiting, to the extent possible, the robot charging station. Our approach is based on building specialized, hardware acceleration units on a Zynq-based FPGA of Xilinx. Moreover, a hardware–software partitioning approach is performed base on the HLS (high-level synthesis) paradigm. In this way, the robot will be able to perform computation-intensive tasks in a power-efficient manner.