Impact-aware robotics boosting efficiency in logistics through predictive models
Maarten Jongeneel defended his PhD thesis at the Department of Mechanical Engineering on November 12th.
Humans naturally interact with objects鈥攍ike grabbing a coffee cup or packing a laptop鈥攖hrough quick, intuitive motions, often involving impacts. Robots, however, avoid impacts to maintain control, resulting in slower, more cautious movements. Impact-aware robotics aims to change this by enabling robots to predict and utilize impacts, enhancing their speed and efficiency. For his PhD research Maarten Jongeneel developed and validated predictive models for impact dynamics, integrated them into a physics engine, and provided a robust dataset to support further advancements. These contributions pave the way for faster, impact-aware robotic applications, especially in high-demand logistics settings.
The tasks mentioned above involve making and breaking physical contact between the hand-held objects and the surrounding environments. Without us realizing it, these interactions often involve impacts, a contact established at nonzero speed. The capabilities of humans to understand, predict, and exploit these impacts allow us to grab and toss objects, speeding up many of our day-to-day tasks. In industry, traditional robotic applications fear impacts as they can lead to vibrations in the robot that complicate the control of the robot. As a result, contact is often established at zero speed, making these systems significantly slower compared to human operators.
Predicting dynamic behavior
Impact-aware robotics is a research field focusing on exploiting intentional impacts with robots to enhance their capabilities. As these impacts often happen in the order of milliseconds, they are too fast to be reacted upon. As a result, one aspect of this research involves predicting the dynamic behavior of the objects and robots as they make impact. These predictions can be used to anticipate impact events in planning and control strategies of the robot. In his research, focused on developing and validating models that can predict this behavior. His work specifically targets logistics applications, where demand for autonomous robotic systems is high due to factors like labor scarcity, an aging population, and high production costs.
High levels of accuracy
He studied contact interactions between a robotic arm, objects such as boxes, and the environment surrounding them (e.g., a conveyor belt). And demonstrated the effectiveness of his approach by combining the developed models into a single physics engine that is used to simulate complex robotic tasks, reaching high levels of accuracy. The vast amount of experimental data collected in this study is shared in a newly proposed database to stimulate further research in impact-aware robotics. Combined, the findings of this thesis provide a solid contribution via validated models capable of predicting the outcome of contacts and impacts, speeding up robotic tasks to enable impact-aware robotics in logistics applications.
Title of PhD thesis: . Supervisors: Associate Prof. Alessandro Saccon, and Prof. Nathan van de Wouw.