Conference Paper

BANSAI: Towards Bridging the AI Adoption Gap in Industrial Robotics with Neurosymbolic Programming

We are happy to announce that our paper “BANSAI: Towards Bridging the AI Adoption Gap in Industrial Robotics with Neurosymbolic Programming” has been accepted at the 57th CIRP Conference on Manufacturing Systems (CMS) 2024.

The full text of the paper is accessible on ArXiv (Link).

The BANSAI workflow models the industrial robot programming process, the roles and involvement of human actors, as well as opportunities for AI assistance.

In the paper, we address the gap between advanced AI/deep learning capabilities in robotics and their limited adoption in industrial robot programming. We propose BANSAI (Bridging the AI Adoption Gap via Neurosymbolic AI), an approach that combines symbolic AI with deep learning to make AI more practical for industrial robotics applications. The key innovation is a dual program representation system that maintains both a traditional skill-based representation for human interaction and robot control, alongside a neural “surrogate” representation for learning and optimization. This allows BANSAI to integrate AI assistance into existing industrial robotics workflows while preserving important properties like human interpretability, safety certification, and integration with factory systems. The approach aims to overcome major challenges that have limited AI adoption in industrial robotics, such as program complexity, physical manipulation requirements, and the need for human oversight and trust.