The lightweight design of electric car body and battery tray is an important way to improve the endurance mileage of electric vehicles, which is a multidisciplinary optimization (MDO) problem with multiple design variables and multiple constraints such as stiffness, modal, and crashworthiness performances, thereby leading it to be inefficient. To improve the MDO efficiency while ensuring the search accuracy, an improved collaborative optimization (CO) framework is proposed in this paper including a system-level optimization, multiple disciplinary subproblem optimizations, and a loop control. The disciplinary subproblem optimizations use the evolutionary optimization algorithm for global search, whose current optimal solutions are combined as the starting point of system-level optimization. The system-level optimization uses the gradient based optimization algorithm for local search, whose current optimal solution is divided into the starting points of the disciplinary subproblem optimizations. The above process is repeated until the convergence condition of the loop control is satisfied. The loop control is mainly used for the data transfer and convergence judgment between the disciplinary subproblem optimizations and the system-level optimization. The convergence and the efficiency of improved CO framework are verified by solving a simple MDO problem. In solving the MDO of electric car body and battery tray with all constraints satisfied, the iteration number and the computational time of improved CO framework are respectively reduced by 69.3% and 60% than those of original CO framework.