TacDiff: Designing Tactile-based Robots via Differentiable Simulations

This fellowship aims to revolutionise robotic dexterity by co-optimising the design, sensing and control aspects of robot manipulators. Its focus lies in developing differential simulations for tactile-based robot manipulator designs that play a crucial role in enhancing robot dexterity and facilitating safe interaction with the unstructured environment. Tactile robots offer the potential for precise manipulation and interaction with the environment. However, current approaches lack integration and efficiency, often leading to isolated development of manipulator morphology, tactile sensing mechanisms, and control algorithms. Further, manual and time-consuming hardware design processes hinder scalability and optimisation across various components, and ensuring the manipulator’s adherence to desired specifications proves challenging due to intricate interplays between robot design, manufacturing constraints, and the control algorithm. The need for a cohesive, integrative approach is evident, where manipulator design, sensing capabilities, and control strategies are co-optimised to achieve enhanced dexterity and adaptability. This fellowship aims to establish the UK's leadership in tactile robotics. Specifically, the objectives of this fellowship are: To establish the mathematical framework for tactile robot parameterisation, enabling the co-design of structure, sensing, and robot behaviours, and thus facilitating a comprehensive design space for advanced tactile robots. To develop differentiable simulations that accurately predict a robot’s interactions with its environment while automatically refining its sensing, structure, and control configurations based on its interaction history with the environment. To enhance the performance of tactile robots and their simulations by integrating real-world data and implementing bidirectional learning between real and simulated environments. This fellowship offers a range of applications with significant impact. In industry, the fellowship will transform sectors requiring precise, adaptive manipulation, such as manufacturing, logistics, and healthcare. Partnerships with companies like Unilever, Ocado, and Shadow Robot Company will facilitate the integration of cutting-edge tactile sensors into commercial robotic systems, enhancing performance and efficiency. This research will drive innovation in industrial automation and position the UK as a leader in robotics technology. The societal impact includes advancing robotic capabilities where precision and adaptability are essential, such as in healthcare for the elderly and disabled, and promoting sustainability by reducing waste through better logistics handling. The fellowship will address pressing societal challenges, contributing to a future where robotics significantly improves everyday life. Educationally, the fellowship will nurture the next generation of roboticists through TaRoSim, an educational toolbox designed to make advanced tactile robotics accessible to students in high school and higher education. By offering interactive resources and hands-on learning experiences, TaRoSim will inspire future innovators and build a skilled workforce ready to drive future advancements in robotics. Outreach initiatives, including workshops at events like Girls into Electronics, will ensure diverse engagement, empowering future leaders in the field and solidifying the UK’s position as a leader in robotics education and research.