Reinforcement Learning at Design of Electronic Circuits: Review and Analysis

Electronic circuit design is a complex, complicated and iterative process, aiming to produce a suitable topology and output parameters considering a predefined specification. The designer has to consider a wide variety of possible choices to obtain the optimal circuit solution. Once the circuit is created, the designer has to figure out the floor plan of its blocks, the placing and wiring/routing the components on printed circuit board (PCB) or on chip by avoiding collisions and taking into account various constraints. Such a repetitive process without automated steps is time, effort and resources consuming. This is the reason for the recent research interest in applying new techniques and methods supporting decision making as reinforcement learning (RL) and deep reinforcement learning (deep RL). Thus, the aim of the current investigation is to summarize and analyze contemporary scientific achievements regarding the benefits of implementing RL and deep RL in the electronic circuit design process and highlighting emerging trends and future research directions.