Multi-FLEX: An Automatic Task Sequence Execution Framework to Enable Reactive Motion Planning for Multi-Robot Applications

TL;DR

Multi-FLEX addresses the challenge of coordinating multiple industrial robots under variability by integrating an online task planner with a reactive motion planner. It introduces task decomposition (Idle, RM, NRM, NMA) and task bundling, plus a Motion Coordinator to resolve overlaps, all interfacing with a dynamic road-map–based Global Planner and a fast Local Planner that uses a QP for collision avoidance. The approach is demonstrated on a two-robot deburring cell, achieving successful task completion, collision avoidance, and significant productivity gains over a single offline-planned cell. This framework offers a practical, flexible path to deploying multi-robot systems in real-world manufacturing settings where tasks and timings vary dynamically.

Abstract

In this letter, an integrated task planning and reactive motion planning framework termed Multi-FLEX is presented that targets real-world, industrial multi-robot applications. Reactive motion planning has been attractive for the purposes of collision avoidance, particularly when there are sources of uncertainty and variation. Most industrial applications, though, typically require parts of motion to be at least partially non-reactive in order to achieve functional objectives. Multi-FLEX resolves this dissonance and enables such applications to take advantage of reactive motion planning. The Multi-FLEX framework achieves 1) coordination of motion requests to resolve task-level conflicts and overlaps, 2) incorporation of application-specific task constraints into online motion planning using the new concepts of task dependency accommodation, task decomposition, and task bundling, and 3) online generation of robot trajectories using a custom, online reactive motion planner. This planner combines fast-to-create, sparse dynamic roadmaps (to find a complete path to the goal) with fast-to-execute, short-horizon, online, optimization-based local planning (for collision avoidance and high performance). To demonstrate, we use two six-degree-of-freedom, high-speed industrial robots in a deburring application to show the ability of this approach to not just handle collision avoidance and task variations, but to also achieve industrial applications.

Figures (12)

• Figure 1: Simplified representation of the Multi-FLEX framework, shown as green blocks. Based on the feedback from the Motion Coordinator, the Task Planner sequentially sends sub-task information to the Motion Coordinator. The Motion Coordinator uses this information to generate robot occupancy information (used for collision avoidance) for the online reactive motion planner.
• Figure 2: An example of a task dependency graph. Multi-FLEX can dynamically assign parallel tasks to robots as they become available, but waits until sequential tasks are completed before moving on.
• Figure 3: Categories of sub-task types during robot task execution. 1) Idle (robot maintains its pose for a given duration) 2) Reactive motion, or RM (online replanning with reactive collision avoidance, e.g., the final approach before picking a part) 3) Non-reactive motion, or NRM (path/speed constrained motion without reactive collision avoidance), and 4) Non-motion action, or NMA (e.g. gripper open/close to grasp the part shown in the figure).
• Figure 4: Illustration of goal overlap between two robots, labeled R1 and R2. Both robots are assigned to pick an object from the part feeder (in the center, between the robots), thus creating a goal overlap. The Motion Coordinator de-conflicts the goal overlap situation by assigning priority and manipulating the occupancy information artificially.
• Figure 5: Illustration of voxel-based occupancy reservation for RM, NRM, and NMA, shown in green. In Figures (a) and (c), the robot occupancy is reserved only at the goal pose. In Figure (b), the entire point-to-point path is reserved. So, specific paths/motions required by the application (NRMs) are protected from interruption, while other motions (RMs) are left to be dynamically updated during the motion.