Safety Query: 3D Collision Mass Maps for Human-Robot Interaction

Institute

Professur für Cyber Physical Systems (TUM-CIT)

Type

Master's Thesis /

Content

experimental / constructive /  

Description

This project aims to extend Collision Mass Maps to full 3D workspaces and redundant robotic manipulators. The goal is to develop a data-driven model that predicts the effective robot mass based on 3D collision location, robot configuration, and possibly collision direction. The extended map should enable more accurate safety assessment for collaborative robots operating in complex workspaces.
A main challenge is the efficient collection of informative collision measurements, since exhaustive sampling in high-dimensional spaces is impractical. Therefore, the project will investigate uncertainty-aware learning methods, such as Gaussian process regression, and active sampling strategies, such as Bayesian optimization, to construct the extended CMM with a limited number of measurements. The resulting framework will be evaluated on a real robotic platform.

This project requires interaction with robotic hardware and therefore regular on-site presence.

Requirements

1. Programming in Python/C++

2. Basic robotics and robot kinematics knowledge

3.Self-motivated

Possible start

18.05.2026

Contact

Yue Zhang
yue22.zhangtum.de

Announcement