Robot Cell Modeling via Exploratory Robot Motions: A Novel and Accessible Data-Driven Approach

TL;DR

The paper tackles the challenge of generating collision-free robot motions in cluttered, often-modified robot cells by proposing a data-driven approach that uses only the robot’s internal joint encoders during exploratory motions to build a conservative, mesh-based model of the obstacle-free space. The method computes per-link swept volumes from recorded joint trajectories, optionally decimates them, and derives an obstacle representation VO by subtracting these volumes from a bounding workspace, enabling integration with standard planning and control frameworks. A cube-shaped exploration tool and a robot-agnostic design speed up exploration and broaden applicability, demonstrated on a KUKA LBR iisy in a pick-and-place task with total modeling time under eight minutes. The approach eliminates CADs and external sensors, reduces costs, and supports flexible production lines, while clearly outlining limitations for static environments and non-real-time processing. Overall, it provides a practical, accessible pathway to collision-free robot operation in dynamic industrial contexts where rapid reconfiguration is common.

Abstract

Generating a collision-free robot motion is crucial for safe applications in real-world settings. This requires an accurate model of all obstacle shapes within the constrained robot cell, which is particularly challenging and time-consuming. The difficulty is heightened in flexible production lines, where the environment model must be updated each time the robot cell is modified. Furthermore, sensor-based methods often necessitate costly hardware and calibration procedures and can be influenced by environmental factors (e.g., light conditions or reflections). To address these challenges, we present a novel data-driven approach to modeling a cluttered workspace, leveraging solely the robot internal joint encoders to capture exploratory motions. By computing the corresponding swept volume (SV), we generate a (conservative) mesh of the environment that is subsequently used for collision checking within established path planning and control methods. Our method significantly reduces the complexity and cost of classical environment modeling by removing the need for computer-aided design (CAD) files and external sensors. We validate the approach with the KUKA LBR iisy collaborative robot in a pick-and-place scenario. In less than three minutes of exploratory robot motions and less than four additional minutes of computation time, we obtain an accurate model that enables collision-free motions. Our approach is intuitive and easy to use, making it accessible to users without specialized technical knowledge. It is applicable to all types of industrial robots or cobots.

Figures (9)

• Figure 1: Swept volume (in orange) of a KUKA LBR iisy robot (left: front view, right: back view). The exploration of the free workspace has been performed through hand guidance with a cube-shaped exploration tool.
• Figure 2: The proposed pipeline consists of 4 steps. First, exploratory robot motions sweeps the free space of the constrained robot cell. The recorded joint trajectories ${\bm{q}\left( t \right)}$ are utilized to compute the robot link poses and, thus, the corresponding link swept volumes ${V_{i}}$. These 3D meshes can optionally be decimated to obtain a simplified volume ${\widetilde{V}_{i}}$ while preserving the overall shape. We obtain a representation of the unexplored and potentially occupied space ${V_{O}}$ by carving out the link swept volumes from a bounding volume that covers the entire robot workspace. Steps 1 -- 3 can be optionally repeated in additional exploration sessions to improve the representation of ${V_{O}}$. The 3D mesh associated with ${V_{O}}$ can subsequently be utilized within established methods for a collision-free trajectory planning and control.
• Figure 3: Example scenario where a gripper moves along a static metal bar, positioning its fingers on either side of the obstacle (left). Using the gripper's convex hull (or bounding box) would incorrectly label the volume between the two gripper fingers as obstacle-free, despite being occupied (right).
• Figure 4: Results obtained with step 2 and 4 of our proposed pipeline for a planar robot (in red) with three non-static links.
• Figure 5: Swept volumes of a KUKA LBR iisy 3 for three different motions: a rectangular (left) and semi-circular (middle) path for a gluing task, and the exploration of a shelf (right) for an inspection task.