Exploring the Development of Intelligent Designer Assistance System Using Deep Reinforcement Learning

Abstract

While Artificial Intelligence (AI) models have the ability to provide intelligent designer support, there are drawbacks as well. These include the possibility of homogenizing design outcomes, overreliance on predefined design parameters, difficulties in accommodating changing user needs, and potential limitations in capturing nuanced design preferences. In this research, exploring the Development of Intelligent Designer Assistance System Using Deep Reinforcement Learning (ED-IDAS-DRL-DKNN) is proposed. Initially the input data is collected from the UrbanScene3Ddataset.The input data’s are fed to pre-processing using Low-Pass Virtual Filtering (LPVF).LPVF is used to clean the data. Then, the pre-processed data is fed to Deep Kronecker Neural Networks (DKNN) as it improves the efficiency of architectural space design and reduces the cost. In order to improve the efficiency and reduce the cost accurately Deep Kronecker Neural Networks is optimized using the Golden Search Optimization (GSO). The proposed method is implemented in Python. The efficiency of the proposed ED-IDAS-DRL-DKNN approach is evaluated using a number of performance criteria including Accuracy, precision, design error and model fitting degree. The proposed ED-IDAS-DRL-DKNN methods attains 16.33%, 35.42% and 28.27% higher precision, 19.36%, 23.42% and 35.42% higher accuracy compared with existing methods such as "Towards intelligent design": An AI-based fashion designer using generative adversarial networks assisted by sketch and rendering generators (AIFD-SRG-GAN), "Exploration of the intelligent-auxiliary design of architectural space using artificial intelligence model" (IADAS-AIM-ANN), and "Intelligent designs in Nano photonics from optimisation towards inverse creation" (IDNP-OTIC-DNN), respectively.