Robelix concept for cable harnessing automation - Neura Robotics Challenge

Описание: This project was presented at the Neura Robotics Challenge 2025, where it was evaluated by a jury of renowned robotics experts. After careful assessment, it was awarded second place.

Description:

In this project, we aim to facilitate the automation of complex manufacturing applications using a versatile mobile manipulator that can dynamically work at different work stations in the factory. The core application is an industrial cable routing application on a large cable harness board. We demonstrate the capabilities of our framework to handle deformable objects with efficiency and reliability. In addition to this core application, we demonstrate the versatility of the system by showing an additional assembly application involving contact situations that require force control and vision. Both applications were implemented with the use of re-usable robot skill libraries that are defined with our constraint-based, reactive, sensor-based framework.

Our solution consists of the MAV platform working together with the MAiRA arm to deal with larger assembly boards and larger cable racks. Similarly to human operators, we use force-based skills where the robot actively keeps the cable under a desired tension while manipulating it. These force-based skills require low-level control (i.e. closed-loop control at 1kHz) that is done through the C++ interface. The realtime controller is automatically generated from higher-level task specifications through our expressiongraph-based Task Specification Language eTaSL.

Our proposed solution uses MAV’s mobility while compensating for its slower dynamics using the faster dynamics of the MAiRA robot arm. This enables us to use MAV+MAiRA simultaneously and together to execute the sensor-based skills necessary to perform the cable routing in a human-like way. We employ a modular control architecture for constraint-based, reactive, and sensor-based control based on our open-source software eTaSL.