Robot throwing has emerged as a promising technique to enhance flexibility and efficiency in object transportation tasks, particularly in factory settings. To further increase the throwing system’s throughput, we explore the multiobject multi-target throwing. In this work, we propose a unified planning framework for multi-object multi-target throwing. The key challenge of multi-target multi-object throwing is to generate a smooth throwing motion efficiently. To address this, we model the object’s inverted flying dynamics and the robot’s kinematic feasibility. By incorporating efficient solution matching and filtering strategies, our method enables the generation of multi-object multi-target throwing motions in a short amount of time (less than 5 ms) within the reachable space, and the whole trajectory duration presents 29% time-consuming decreasing compared to the naive approach. Due to the computational efficiency of our approach, we demonstrate the adaptability and reactivity of our system under disturbances, allowing for real-time re-planning and alternative solution generation.
Zhengming Zhu (Wed,) studied this question.