System and method for autonomous printed circuit board design using machine learning techniques

A non-transitory computer-readable storage medium having stored thereon instructions, which when executed by a processor result in one or more operations, the operations comprising:

• receiving, using a processor, an electronic design;
• analyzing the electronic design using a reinforcement learning agent, wherein analyzing includes analyzing a feature vector including both images and scalars and wherein the images include electronic design representations and the scalars are selected from the group consisting of:
• printed circuit board dimensions, wirelength, number of nets, number of packages, and discrete components;
• recommending a first action wherein the first action includes at least one of a place agent action, a via agent action, or a route agent action;
• updating the electronic design based upon, at least in part, the first action to generate an updated electronic design;
• analyzing the updated electronic design using the reinforcement learning agent, wherein analyzing includes analyzing a feature vector including both images and scalars and wherein the images include electronic design representations and the scalars are selected from the group consisting of:
• printed circuit board dimensions, wirelength, number of nets, number of packages, and discrete components;
• recommending a second action wherein the second action includes at least one of a place agent action, a via agent action, or a route agent action;
• and
• updating the updated electronic design based upon, at least in part, the second action to generate a second updated electronic design.

A system, comprising:

• at least one processor configured to receive an electronic design, the at least one processor further configured to analyze the electronic design using a reinforcement learning agent, wherein analyzing includes analyzing a feature vector including both images and scalars and wherein the images include electronic design representations and the scalars are selected from the group consisting of:
• printed circuit board dimensions, wirelength, number of nets, number of packages, and discrete components, the at least one processor further configured to recommend a first action, wherein the first action includes at least one of a place agent action, a via agent action, or a route agent action, the at least one processor further configured to update the electronic design based upon, at least in part, the first action to generate an updated electronic design, the at least one processor further configured to analyze the updated electronic design using the reinforcement learning agent, wherein analyzing includes analyzing a feature vector including both images and scalars and wherein the images include electronic design representations and the scalars are selected from the group consisting of:
• printed circuit board dimensions, wirelength, number of nets, number of packages, and discrete components, the at least one processor further configured to recommend a second action wherein the second action includes at least one of a place agent action, a via agent action, or a route agent action, the at least one processor further configured to update the updated electronic design based upon, at least in part, the second action to generate a second updated electronic design.