Computational Design Synthesis of Passive Dynamic Brachiating Robots

Topologic configurations of robotic systems are usually designed by hand by a human designer, utilizing experience and intuition. Only a few alternative configurations are often evaluated through simulation involving interactive refinement of the simulation model for both computational efficiency and accuracy. This project aims to develop automated design methods for such systems to generate valid, optimized and potentially innovative solutions compared to human-based design. The approach to automate the design process computationally combines graph grammars for generative design and optimization to produce a range of new, valid robot configurations. Once a new configuration is generated, automated simulation model building, dynamic simulation and evaluation of the new configuration are carried out. The performance of the whole process strongly depends on the efficiency of the dynamic simulation. This removes the human from the loop enabling a wider range of robotic configurations to be produced. While use of design automation increases in industry, simulation-driven design automation is still emerging and has great potential to help designers achieve more innovative, reliable and cost-effective solutions.