Integrating data to fight cancer: Sharing ideas and best practice
5 December 2024
By Yue Xie, Marie Sklodowska-Curie Future Roads Fellow, Bio-Inspired Robotics Laboratory, Department of Engineering
20 June 2024
Embodied artificial intelligence (EAI) brings AI, robotics, and bioengineering together. Embodiment in artificial intelligence is foundational for the mechanical execution of tasks and to achieve a higher order of situational and adaptive intelligence.
EAI is rooted in the idea that true intelligence transcends information processing to encompass physical and social interactions. Interacting with the physical world requires a body. Embodiment enables tangible robots to understand and adapt to complex environment.
Within the field, evolutionary soft robotics merges evolutionary algorithms and the analytical benefits of AI with the mechanical sophistication of robotics. This niche - which is my current area of interest - focuses on the harmonious integration of sensing, mechanics, and control within soft structures while highlighting the complex interplay between embodied intelligence and artificial constructs.
Evolutionary soft robotics enables the development of robots that are flexible, resilient and capable of self-improvement over time by integrating evolutionary algorithms which simulate the process of natural. This means that they can adapt their behaviour based on environmental feedback without human intervention. Such advanced robots with EAI that could one day be part of our everyday lives.
Sharing knowledge
The field of EAI is emerging, making it extremely beneficial to share knowledge across the community to drive progress. Using funding from the Accelerate Programme for Scientific Discovery and the Cambridge Centre for Data Driven Discovery (C2D3), I organised a workshop on 6 March 2024 exploring the intersection of EAI and Evolutionary Soft Robotics.
There were 25 participants, including PhD students from Cambridge and EPFL, early-stage researchers from Cambridge, EPFL and Edinburgh Napier University, and professors from USA, AU, EU and UK. The funding allowed us to bring together an international field of experts who shared diverse perspectives at the cutting edge of these fields.
The core idea of our workshop was to discuss the application of embodied AI in evolutionary soft robotics and how we can combine ideas and actions, theory and practice to uncover the latent potential of embodied AI, and shape the trajectory of AI-driven robotics.
We focused on leveraging AI and evolutionary strategies to enhance the adaptability, efficiency and functionality of soft robots. We explored several key research directions, including optimising soft robotic designs for diverse environmental conditions and developing more sophisticated control algorithms for real-time adaptability. Speakers from the universities of Bristol, Edinburgh Napier, Vermont and Vrije University Amsterdam as well as Imperial College London, CSIRO and PFL shared their expertise. For example, David Howard of CSIRO and Josh Bongard from the University of Vermont highlighted the integration of AI techniques in the design and optimization of soft robots, with Howard speaking about how AI can enhance the design process, and Bongard detailing how maintaining effective gradient flows in differentiable programming is crucial for optimising soft robot designs in real-time.
The design process of robots and real-time optimisation are critical because they enable the creation of robots that can adapt to changing environments and perform tasks more efficiently. For example, in industrial settings, a robot with an enhanced design process can handle a wide range of products on an assembly line, improving productivity and reducing downtime due to reconfiguration.
The workshop gave me the opportunity to showcase my ideas in front of an engaged audience, while receiving feedback from peers and experts has helped me refine my approach and think more critically. These skills will be essential as I take forward my research in this area . Furthermore, organising and managing an international conference gave me invaluable experience in project management and coordinating logistics, facilitating cross-disciplinary collaboration, and fostering a sense of community among researchers.
My research
My research centres on EAI and more specifically, evolutionary robotics. I’m concerned with how to design a robot’s control centre or ‘brain’ as well as its morphologies. I use AI in every part of my work to build robots. I used deep learning, or reinforced learning to train the control centres of the machines, but I also use it to design morphologies of robots without the need for any human experience. We call this bio inspired robotics design framework.
Currently, I am working on a hand with grippers, the shape of which is inspired by a fish fin. A gripper like ours has many potential industrial and agricultural applications. For example, it could be used to handle and pack fragile items like glassware, or for harvesting fruit and vegetables without bruising or spoiling produce. My team’s work on the fin ray soft gripper, proposes an automated computational design optimization framework that generates gripper diversity to individually grasp geometrically distinct object types based on a quality-diversity approach.
It first discusses a large design space including 28 design parameters for a finger-based soft gripper, including the rarely-explored design space of finger arrangement. Then, a contact-based Finite Element Modelling (FEM) is proposed in a simulation software called SOFA (Simulation Open Framework Architecture) to output high-fidelity grasping data for fitness evaluation and feature measurements. Finally, diverse gripper designs are obtained from the framework while considering features such as the volume and workspace of grippers. This work bridges the gap of computationally exploring the vast design space of finger-based soft grippers, while grasping large geometrically distinct object types with a simple control scheme.
The work advances how computational design can be used in my field of EAI and soft robotics, as well as the frame design space. This field is crucial for applications in healthcare, manufacturing, and agriculture, where robots need precision and adaptability.
The future
The future of embodied artificial intelligence holds transformative potential, extending far beyond the current horizons of robots and AI. It could lead to autonomous agricultural robots that could more efficiently tend to crops, or robots that could navigate hazardous environments to conduct search and rescue operations in the aftermath of natural disasters.
I already have a lot of collaborations planned, as well as upcoming workshops, including our next event at a major computer science conference in Melbourne, We are currently calling for contributions for this event, which aims to fuse theoretical computer science with practical EAI solutions. By bridging these disciplines, we hope to unlock new efficiencies and capabilities, such as leveraging advanced theoretical algorithms could lead to the development of more autonomous, decision-making robots that can adapt and learn from their environments in unprecedented ways.
We imagine a time when robots can help us with everything. One ongoing project in our lab is to create a chef robot to help with making dinner, but robots could help us communicate too. There are so many opportunities and hopefully, the workshop might help make some of them reality. From medical robots with arms that can adapt to surgical tasks and support in varied procedures, to robotic chefs, who knows what we can cook up!.
Learn more about the conference taking place in July 2024. Call for contributions
Yue was awarded funding through the Accelerate-C2D3 funding call. Find out more about the 2024 call for proposals here and apply by 18 September.