This paper proposes a co-worker in-time Parts/tools Delivery to You robot, PaDY, in an automobile assembly line. Since co-worker robots share a workspace, the workers are at risk of colliding with them. Therefore, the safety of the workers has to be considered when operating such robots. In previous studies, the robot generally stops to ensure the worker's safety when a collision is detected, delaying the work process. However, if the robot can avoid the collision and continue to operate smoothly, the work process will not be delayed. Hence, this paper proposes a novel dynamic collision avoidance method for the co-worker robots. In this method, we utilize Markov model and Kalman filter to predict the trajectory of the workers. We define the Time Augmented Configuration-Space as the configuration space with a time axis and use it for obstacle depiction and goal setting. We applied rapidly-exploring random tree search method for trajectory planning in Time Augmented Configuration-Space. Finally, we simulate collision avoidance with workers. The improved collision avoidance mechanism proposed here is a step toward allowing workers and robots to share a workspace, which will enhance productivity in the automobile sector in particular.