Provably constant-time planning and replanning for real-time grasping objects off a conveyor belt

Abstract

In warehouse and manufacturing environments, manipulation platforms are frequently deployed at conveyor belts to perform pick-and-place tasks. Because objects on the conveyor belts are moving, robots have limited time to pick them up. This brings the requirement for fast and reliable motion planners that could provide provable real-time planning guarantees, which the existing algorithms do not provide. In addition to the planning efficiency, the success of manipulation tasks relies heavily on the accuracy of the perception system which is often noisy, especially if the target objects are perceived from a distance. For fast-moving conveyor belts, the robot cannot wait for a perfect estimate before it starts executing its motion. In order to be able to reach the object in time, it must start moving early on (relying on the initial noisy estimates) and adjust its motion on-the-fly in response to the pose updates from perception. We propose a planning framework that meets these requirements by providing provable constant-time planning and replanning guarantees. To this end, we first introduce and formalize a new class of algorithms called constant-time motion planning (CTMP) algorithms that guarantee to plan in constant time and within a user-defined time bound. We then present our planning framework for grasping objects off a conveyor belt as an instance of the CTMP class of algorithms. We present it, provide its analytical properties, and perform an experimental analysis both in simulation and on a real robot.

Provably Constant-time Planning and Replanning for Real-time Grasping Objects off a Conveyor Belt

“…This article is in continuation of our previous work presented in Islam et al (2020) and the contributions 4 and 5 specifically, are the extensions. In addition to these extensions, we provide space complexity analysis of our approach and report detailed preprocessing statistics of our experiments to highlight the improvement over the brute force method.…”

“…In addition to these extensions, we provide space complexity analysis of our approach and report detailed preprocessing statistics of our experiments to highlight the improvement over the brute force method. We also remove one of the assumptions of Islam et al (2020) which says that the environment remains static up to a certain time in execution.…”

Efficient online motion planning method for the robotic arm to pick-up moving objects smoothly with temporal constraints

“…The existing real-time grasping object methods mainly include the visual servo control, 1,2 reinforcement learning, 3 and motion planning. 4–7 The tracking and grasping scheme based on visual servo can solve the problem of grasping random moving targets but cannot deal with the object occlusion and cannot generate smooth joint trajectories. The reinforcement learning methods are not stable in practical application.…”

“…Yang et al 4 used the RRT-Connect method in time configuration space to generate the path to avoid moving obstacles. But the algorithm’s completeness guarantee is weakened because the manipulator is prespecified to grasp the target at the nearest position to the robot base 4-7 and does not carry out trajectory optimization. The joint trajectories are poor smooth, and some waypoints are close to the obstacle, so the robot arm cannot perform the grasping tasks smoothly and safely.…”

“…The reinforcement learning methods 3 are not stable in practical application. The motion planners use sampling-based methods 4 or search-based methods [5][6][7] to find feasibility paths, but they lack back-end optimization to smooth jerky paths because the existing classical trajectory optimization algorithms, like CHOMP 8 and STOMP, 9 are unsuitable for a moving target due to a large amount of calculation. As a result, the lack of smoothness makes the trajectory inefficient and less human-friendly.…”