Siemens Digital Industries Sofware Internship

Reactive Trajectory Planning for Robotic Operations in Tecnomatix Process Simulate

Objective

This internship project, conducted at Siemens Digital Industries Software in Toulouse, France, within the Kineo¹’s Advanced Robotics Kinematics (ARK) team, focuses on integrating a reactive trajectory planning system into Tecnomatix Process Simulate. The goal is to enable robotic arms to dynamically perceive and avoid obstacles in unstructured environments using real-time RGB-D camera data.

Key Components

Architectural Design

A layered architecture was implemented to ensure modularity and performance:

UI Layer (C#/WPF): A user-friendly dialog for configuring and controlling the reactive system, including camera setup, simulation start/stop, and real-time feedback.
Bridge Layer (C++/CLI): Acts as a mediator between managed C# code and native C++ components, using wrappers to handle data transfer and command execution.
Core Engine (Native C++): Manages the simulation loop, event-driven point cloud capture, and integration with Kineo SDKs for collision detection.

Data Acquisition and Processing

RGB-D cameras simulate sensor input in Process Simulate. Depth buffers are captured at timed intervals during simulation:

Point Cloud Generation: Converts 2D pixels to 3D points using camera intrinsics (focal length, principal point) and applies camera-to-world transformations.

Two-Stage Filtering Algorithm

To isolate dynamic obstacles:

Stage 1: Robot Geometry Filtering: Removes points within the robot’s bounding boxes (e.g., links of a UR5 arm) to eliminate self-occlusion.
Stage 2: Static Environment Filtering: Uses voxel grids to subtract known static elements like bins or floors.

Collision Detection

Integrates Kineo Collision Detector to check the robot’s trajectory against the filtered point cloud, halting simulation upon potential collisions.

Results

Model Integration and Validation

The system was tested in a bin-picking scenario. It successfully detects unexpected obstacles, isolates them via filtering, and prevents collisions by stopping the robot.

Impact and Future Work

This framework advances reactive robotics in simulated environments, improving safety and adaptability in manufacturing. It reduces reliance on predefined paths and supports human-robot collaboration. Future enhancements include full autonomous path replanning (integrating “plan and act” phases), multi-camera support, hardware-in-the-loop validation, and optimizations for denser point clouds.

Giulio Carpi