Abstract: In the manufacturing environment, there are two problems in the application of energy saving trajectory planning for industrial robot:one is that the robot dynamic parameters are unknown, and the other is that the existing energy saving trajectory planning methods can't guarantee the stability of the results. Therefore, an energy saving trajectory planning for industrial robot in manufacturing environment is proposed, including approximate dynamic identification based on parallel BP (backpropagation) neural networks, and energy saving trajectory solution based on convex optimization (CO). Taking UR3 robot as the experimental platform, the RMSE (root mean squared error) of the approximate dynamic model can converge to 2.05×10^-3 N. m, and the solution stability of convex optimization based trajectory planning is better than the existing parametric trajectory planning. The experimental results show that the proposed energy saving trajectory planning scheme can deal with the problem of unknown robot dynamic parameters in the manufacturing environment, while ensuring the stability of the trajectory planning results, so it is more applicable for industrial robots in the manufacturing environment.