Predictive Collision Detection in Dynamic Environments

Gabriel Streitmatter

Authors:  Gabriel Streitmatter, Dr. Gloria Wiens

Faculty Mentor:  Dr. Gloria Wiens

College:  Herbert Wertheim College of Engineering


Robots working within dynamic environments must be able to look ahead at their trajectories and identify potential collisions with their ever-changing environment. This is of particular interest in human robot collaboration (HRC) in which robots must interact with dynamic and possibly unpredictable humans safely. In this paper, an algorithm is developed to evaluate a robot’s trajectory, evaluate the dynamic environment that the robot operates in, and predict collisions between the robot and the environment. The algorithm is designed to execute quickly to ensure applicability during online execution of the robot’s tasks. The algorithm takes as input the joint motion data defining the robot’s trajectory and constructs a sweep of the robot’s position throughout time by using a representative set of poses of the robot throughout the trajectory and interpolating the robot’s path between them. Coons patches are leveraged to approximate the robot’s position throughout time between the set of know poses. The algorithm creates a similar sweep for any obstacles being tracked in the operating environment. The sweeps are compared to identify collisions. This algorithm is proposed as a “watchdog” collision detector to assist the “robot action” focus of a larger collaborative effort to develop robust HRC techniques.

Poster Pitch

Click the video below to view the student's poster pitch.


Click the image to enlarge.