Adaptive Mechanical Systems for Physical Embodied Intelligence Robotics

点击翻译按钮获取中文摘要

In the last few years, robotics has achieved a remarkable breakthrough in human society. Robots are collecting data on Mars, delivering food to your hotel room or medicine in hospitals, exploring the oceans, and manufacturing a seemingly endless range of products. Tightly connected to robotics, Industry 4.0 and Artificial Intelligence (AI) are the focus of innumerable newspaper articles and the topic of many public policies. Yet, the mechanical design of modern robotic systems is often neglected and designers typically rely on complex control strategies (including AI) as well as sophisticated arrays of sensors and actuators. This application argues that this approach is often incomplete in the sense that it ignores the fact that mechanical systems can also solve problems by themselves. The smartest AI without a similarly smart mechanical embodiment is arguably nothing more than a brain in jar. In this research program, mechanical systems will be designed that complement AI or similarly high-level control systems with a physical embodied intelligence counterpart. The program stems from the applicant's earlier efforts in developing mechanically shape changing devices, and proposes to extend the discovered principles to exciting applications along three main research thrusts: variable stiffness robotic hands, adaptive mechanical walking machines, passive shape morphing wings. The first research thrust focuses on self-adaptive robotic fingers that are even "more" underactuated than usual. The applicant and his group have designed novel robotic fingers which are completely passive: the deformation of these fingers and the forces they produce are generated in reaction to contacts on objects and carefully controlled with the design parameters of the fingers. In this application and following the practical experimentation of these prototypes, we are proposing to develop variable stiffness versions of the fingers where one can switch from a very soft and gentle gripper to a stiff and precise one by varying its kinematic structure. The second research thrust deals with another limb of interest in robotics, namely the leg. In recent works, a self-adaptive robotic leg was demonstrated that could react to an unintended collision with an obstacle by changing its kinematics and produce a motion overcoming that obstacle. In this application, we propose to create a complete mechanically intelligent walking robot. Finally, the last research thrust applies lessons from the previous works to morphing wings. We propose to investigate a mechanical scheme where the shape changing of the wing is at least partially caused by aerodynamic forces. This program will have vast implications for the development of AI controlled robotic systems and for Canadian companies active in mobile robotics, autonomous manipulation, and aerospace by providing efficient robotic solutions with a true mechatronic intelligence, i.e. embodied both mechanically and electronically.