Benchmarking Sim2Real Gap: High-fidelity Digital Twinning of Agile Manufacturing

Recently, the book chapter Benchmarking Sim2Real Gap: High-fidelity Digital Twinning of Agile Manufacturing has been accepted for publication by CRC Press, Taylor & Francis Group. It explores the technologies driving the development of digital twins for agile manufacturing in robotic automation, as researched by IMR Ltd, Ireland.

Figure 1. Digital twin technologies for agile robotic automation

This chapter emphasizes the transfer of trained policies and process optimizations from simulated environments to real-world applications, leveraging advanced techniques such as domain randomization, domain adaptation, curriculum learning, and model-based system identification.

Figure 2.Sim2real transfer approaches for agile manufacturing automation; (a) represents Domain Randomization (b) shows Domain Adaptation (c) illustrates Curriculum learning (d) depicts Model based System Identification.

Additionally, the chapter reviews industrial manufacturing automation scenarios, including bin-picking, part inspection, and product assembly, within WP6 of the Coresense project under Sim2Real conditions.

Figure 3. Industrial bin-picking task performed under simulated and real-world domain (a-d) show picking, aligning, repositioning, and placement of cubes into chambers within AI-enhanced simulated settings, (e-h) illustrates similar actions over real robotic setup under semi-stochastic conditions.

The performance of digital twin technologies in these scenarios is evaluated using practical metrics such as data latency, adaptation rate, and simulation fidelity, providing a comprehensive assessment of their effectiveness and potential impact on modern manufacturing processes.

You can access to the book chapter at https://arxiv.org/pdf/2409.10784.

Related Posts
EU Logo