IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 3, NO. 3, SEPTEMBER 1998 225 The Robotic Interception of Moving Objects in Industrial Settings: Strategy Development and Experiment Damir Hujic, Associate Member, IEEE, Elizabeth A. Croft, Member, IEEE, Gene Zak, Robert G. Fenton, James K. Mills, Member, IEEE, and Beno Benhabib, Member, IEEE Abstract— A novel active prediction, planning, and execution (APPE) system is presented herein for the robotic interception of moving objects. The primary feature of the proposed APPE sys- tem is the ability to intercept the object at an optimal rendezvous point, anywhere along its predicted trajectory, within the robot’s workspace. For the interception of objects in industrial settings, the motion of which allows long-term predictability, this feature is a significant improvement over earlier APPE systems. These could only select a rendezvous point among a few nonoptimal interception points considered. An APPE system’s objective is simply to move the robot to the earliest pregrasping location. A fine-motion tracking algorithm can take over the motion control at that point, utilizing proximity sensors mounted on the robot’s end-effector. This approach eliminates the necessity of tracking the motion of the object, as required by conven- tional tracking-based techniques, where the distance between the robot’s end-effector and the object is reduced continuously. In this paper, the proposed APPE system is first briefly introduced, and its individual modules are thereafter discussed in detail. Simulation and experimental results are presented in support of the developed optimal-interception strategy. Index Terms—Active prediction, planning, and execution sys- tem, robotic interception. I. INTRODUCTION A KEY FEATURE OF intelligent robotic systems is the ability to perform autonomously a multitude of tasks without complete a priori information, while adapting to continuous changes in the working environment. An important problem in this field is the robotic interception of moving objects. A common approach to the object-interception prob- lem is the utilization of a prediction, planning, and execution (PPE) strategy [1], [2]. In a PPE strategy, the motion of Manuscript received November 1, 1996; revised Sptember 19, 1997. Recommended by Technical Editor T. J. Tarn. This work was supported by the Natural Sciences and Engineering Research Council of Canada. The work of E. A. Croft was supported by the Canadian Federation of University Women through the Margaret McWilliams Pre-doctoral Scholarship. D. Hujic was with the Computer Integrated Manufacturing Laboratory, Department of Mechanical and Industrial Engineering, University of Toronto, Toronto, Ont., M5S 3G8 Canada. He is now with Celestica Inc., North York, Ont., M3C 1V7 Canada. E. A. Croft is with the Industrial Automation Laboratory, Department of Mechanical Engineering, University of British Columbia, Vancouver, B.C., V6T 1Z4 Canada. G. Zak, R. G. Fenton, J. K. Mills, and B. Benhabib are with the Computer Integrated Manufacturing Laboratory, Department of Mechanical and Indus- trial Engineering, University of Toronto, Toronto, Ont., M5S 3G8 Canada. Publisher Item Identifier S 1083-4435(98)06943-9. an object through a robot’s workspace is predicted. Robot motion to intercept the object is then planned and executed. This approach can be used in an “active” mode (APPE), e.g., [3], where the three stages may be repeated as necessary to ensure the successful completion of the interception task. In this context, a novel APPE system developed and implemented in our laboratory will be described in this paper. APPE-based approaches constitute an alternative to tracking-based techniques, which essentially minimize the difference between the state of the robot’s end-effector and the state of the moving object, [4], [5]. The principal advantage of APPE systems over tracking-based systems is their ability to find an optimal solution to the interception-point planning problem. However, most APPE-based techniques reported in the literature target nonindustrial settings. They normally sacrifice time optimality in favor of a guarantee of interception, either for fast-moving objects or for objects of which the Cartesian path topology and/or velocity profiles vary significantly over short periods of time. Therefore, they choose a rendezvous point among a limited number of potential interception points considered, where these usually correspond to the intersection of the (most current) estimated target trajectory with fixed planes (or lines in the planar cases). In this paper, the proposed novel technique allows the interception of the moving object at the earliest possible time, given the constraints of the robot’s motion capabilities. Namely, the interception is not restricted to a choice among a few potential points, as with other APPE systems, but targeted toward the selection of the best rendezvous point anywhere on the target’s predicted trajectory. Prior to the description of our research results, some APPE systems are briefly reviewed below. Hove and Slotine [6] used a hybrid PPE-tracking system for robotic ball catching. The PPE approach is used to plan the initial motion of the robot. A point on the ball’s path (parabolic in nature) closest to the initial position of the end-effector is chosen as the first potential rendezvous point. The robot is immediately sent to this point. A tracking strategy takes over the control of the robot once the ball passes the first potential rendezvous point. Thus, in this paper, the PPE technique is utilized to improve upon the principal tracking-based method. Andersson [3] used an APPE approach in the development of a robotic ping-pong player. Based on the predicted motion of the ball, three possible interception points are identified 1083–4435/98$10.00  1998 IEEE