Generating Whole-Body Avoidance Motion through Localized Proximity Sensing

TL;DR

This work addresses safe robotic manipulation in unstructured environments when full surface sensorization is impractical. It introduces a proximity-sensing pipeline using sparse multi-zone ToF sensors to build an environment point cloud, filters out the robot body with an AABB-tree-based mesh check, and computes the minimum distance between the environment and the robot mesh to drive a high-priority whole-body avoidance task. The control framework employs a two-layer task-priority scheme where safety (obstacle avoidance) takes precedence over goal-reaching, enabling reactive motions on any point of the robot surface via virtual control inputs. Experimental validation on a UR10e with partial sensor coverage demonstrates improved distance margins (up to 100 mm) and competitive performance against sensor-only baselines in static and human-robot interaction scenarios, highlighting reduced sensorization requirements and robust safety. The approach offers a practical path to safer collaboration by extending reactive avoidance to non-sensorized links and operating without external sensing architectures.

Abstract

This paper presents a novel control algorithm for robotic manipulators in unstructured environments using proximity sensors partially distributed on the platform. The proposed approach exploits arrays of multi zone Time-of-Flight (ToF) sensors to generate a sparse point cloud representation of the robot surroundings. By employing computational geometry techniques, we fuse the knowledge of robot geometric model with ToFs sensory feedback to generate whole-body motion tasks, allowing to move both sensorized and non-sensorized links in response to unpredictable events such as human motion. In particular, the proposed algorithm computes the pair of closest points between the environment cloud and the robot links, generating a dynamic avoidance motion that is implemented as the highest priority task in a two-level hierarchical architecture. Such a design choice allows the robot to work safely alongside humans even without a complete sensorization over the whole surface. Experimental validation demonstrates the algorithm effectiveness both in static and dynamic scenarios, achieving comparable performances with respect to well established control techniques that aim to move the sensors mounting positions on the robot body. The presented algorithm exploits any arbitrary point on the robot surface to perform avoidance motion, showing improvements in the distance margin up to 100 mm, due to the rendering of virtual avoidance tasks on non-sensorized links.

Figures (7)

• Figure 1: 3D model of a robotic manipulator partially covered with multi-zone proximity sensors. On the left side of the picture is highlighted an obstacle detected by one of the sensors. Despite being closer to the second, non-sensorized link of the robot, a traditional approach would use the position of the closest proximity sensor as control point to drive the obstacle avoidance behaviour. Our solution exploits knowledge of the robot's geometry to define a dynamic avoidance task that acts on an arbitrary point over the robot's surface, regardless of its sensorization.
• Figure 2: Simplified 3D description of the mesh removal algorithm. A ray is traced between a proximity sensor and all its sampled measurements, expressed as 3D points. If the ray intersects the robot's model, the measurement is compared with the intersection point and if the distance is lower than a specified threshold, the measurement is discarded (1). The point (2) instead is an observation of an obstacle, and the corresponding ray does not intersect with the robot model, therefore the measurement is valid.
• Figure 3: Visual comparison of the whole-body minimum distance pair algorithm and its counterpart based on sensors mounting points. The white point cloud represents a downsampled model of the robot computed by its visual meshes (a) and (c) and proximity sensors locations in (b) and (d). The blue cloud represents an obstacle resembling our proposed experimental configurations. The red arrow is the vector that has the pair of closest points as its origin and tip.
• Figure 4: Experimental setup configurations. (a) A static planar obstacle is positioned in front of the robot, to evaluate algorithm's behavior on the uncovered flange. (b) Another planar object is placed besides the robot, to evaluate the algorithm's behavior on the uncovered links. (c) Human-robot Collaboration mockup. A toolbox is placed on the left or right of the robot base depending on the trial configuration. A car door is positioned in front of the robot's workspace, being the target for the user's actions.
• Figure 5: Comparative analysis between the presented whole-body algorithm and the baseline during the front plane test.