Towards Automated Clash Resolution Of Reinforcing Steel Design In Reinforced Concrete Frames And Precast Concrete Sandwich Panels Via Reinforcement Learning And Building Information Modeling

The design of reinforcing steel bars(rebars)is particularly significant to reinforced concrete(RC)structures.The adoption of precast concrete elements(PCEs)are becoming popular in civil infrastructures.Since the quality of connections between adjacent elements determines the structure property of precast structures,the design of rebars in RC elements is a mandatory stage in building construction projects.Generally,a good number of rebars are required by a design code,particularly at member connections.Due to the large number of rebar and the complicated rules for arrangement in each design code,it is impractical,labor-intensive,and error-prone for designers to avoid all clashes(i.e.,collisions and congestions)even using computer software.The building information modeling(BIM)technology has been widely utilized for clash-free rebar designs in the present architecture,engineering,and construction industry.However,most existing BIM-based approaches offer the clash resolution strategy for moving components with an optimization algorithm,and are only applicable to the RC structures with regular shapes.In particular,the optimized path of rebars cannot be adjusted to avoid the obstacles,thus limiting the practical applications.Furthermore,most existing studies lack the learning from design codes and constructability constraints to realize automatic and intelligent arrangement and adjustment of rebars for avoiding the obstacles encountered in complex RC structures.Considering these shortcomings,this paper proposes a framework towards automatic rebar design in RC frame and precast concrete sandwich panels(PCSP)without clashes via multi-agent reinforcement learning(MARL)system with BIM.The main contents are as follows:(1)By treating each rebar as an intelligence reinforcement learning(RL)agent,the rebar design problem is modelled as a path-planning problem of multi-agent system.To carrying out the MARL system,RC members have to be transformed into a suitable digital environment.According to the design codes and constructability constraints,the method for geometry information extraction from the BIM and transformation into the grid environments approximating the geometry of the RC members with known boundary conditions is proposed.Further,the path-planning of multi-agent system is carried out in the grid environment,and the generated results are extracted into BIM.(2)By employing the Q-learning(a model-free reinforcement learning algorithm)as the RL engine,the particular form of state,action,and rewards for the reinforcement MARL for automatic rebar designs considering constructible constraints and design codes is developed.(3)For automatically generating clash-free rebar designs in complex precast concrete sandwich panels from actual engineering,the proposed framework is extended with Generative Adversarial Network(GAN)and Deep Q-Network(DQN).In particular,by using GAN learning from structural design drawings and generating 2D rebar designs,and,the designs are transformed into digital environments for DQN.Next,by employing DQN as the reinforcement learning engine,the generated paths of rebar are modified to avoid the clashes.(4)Comprehensive experiments on three typical beam-column joints,a two-story RC building frame and complex precast concrete sandwich panels from actual engineering were conducted to evaluate the efficiency of the proposed method.The study results of paths of rebar designs,success rates,and average time confirm that the proposed framework with MARL and BIM is effective and efficient.