WEBVTT 00:00:02.487 --> 00:00:09.487 Hi, my name is Connor Heron and today I'll be presenting the work titled The Design and Control of a Structurally Elastic Humanoid Robot. 00:00:09.425 --> 00:00:14.425 Recently, a labor crisis has affected the global market where countries such as the US, Germany. 00:00:14.989 --> 00:00:19.989 South Korea and China need upwards of 5.5% of their population to fulfill their current business demand. 00:00:19.804 --> 00:00:28.804 As the global age is continuing to rise, these costs and healthcare demands are only expected to increase Which drives the demand for technological innovation. 00:00:28.676 --> 00:00:37.676 Humanoid robots are uniquely positioned to address many of these challenges since their human-centric form factor allows them to ergonomically transition to the existing workplace. 00:00:38.303 --> 00:00:47.303 However, it's important to know that a majority of these robots are made via rigid metal components that cost hundreds of thousands of dollars to manufacture. And since they're metal and rigid. 00:00:46.554 --> 00:00:52.554 Doesn't necessarily translate well to safety and spaces collaborating with humans. 00:00:52.551 --> 00:01:04.551 Therefore, this work explores a new humanoid robot design that leverages 3D printing compliant materials offering a dramatic drop in savings, weight, design complexity, and potential for safe contact with humans. 00:01:05.243 --> 00:01:13.243 A good example for why 3D printing technologies are worth investigating is the hydraulic Atlas humanoid robot from Boston Dynamics. 00:01:13.941 --> 00:01:28.941 That utilized metal 3D printing to manufacture several of its high strength linkages. This approach enabled engineers to embed hoseless routing the hydraulic fluid directly into the structural components, which otherwise would have been impossible to manufacture with baseline subtractive manufacturing methods. 00:01:29.067 --> 00:01:40.067 Our design focuses on incorporating compliance directly into the robot design, which is important for legged robots since it protects the hardware from high impact forces from walking. 00:01:39.816 --> 00:01:46.816 As opposed to contemporary approaches that incorporate the elastic elements into the actuator mechanism. 00:01:46.817 --> 00:01:51.817 Pandora is designed to be compliant under load, or in other words, structurally elastic. 00:01:51.879 --> 00:01:57.879 This manufacturing approach dramatically simplifies the robot design, cost, and number of components. 00:01:57.627 --> 00:02:02.627 But introduces control and state estimation challenges that are the focus of this work. 00:02:03.067 --> 00:02:15.067 It's important to keep in mind that to our knowledge, no other robot has been introduced with this design approach while also identifying and addressing these specific challenges, especially at a height of six feet tall for the full design. 00:02:15.817 --> 00:02:25.817 Regarding my methodology, the dissertation is contributes four main bodies of work on the development and control at every layer of the humanoid robot design. 00:02:26.065 --> 00:02:31.065 The first chapter introduces the open source hardware design, which includes the electrical system. 00:02:30.691 --> 00:02:35.691 Network and device synchronization, sensor integration and filtering, and motor control. 00:02:35.504 --> 00:02:46.504 The second chapter focuses on joint control where the actuators must be commanded to track joint trajectories. This is difficult to achieve since there's no good methods for estimating the stiffness properties of 3D printed parts. 00:02:46.380 --> 00:02:52.380 So this work utilizes a robust control approach that treats the unmodeled structural elasticity as disturbance. 00:02:52.691 --> 00:02:59.691 The third chapter focuses on addressing the kinematic challenges where the physical robot and kinematic system did not match during stance transitions. 00:03:00.132 --> 00:03:03.132 When a substantial amount of force is applied on one of the legs. 00:03:02.818 --> 00:03:08.818 Finally, the last chapter introduces a theoretical contribution for planning center of mass trajectories. 00:03:08.377 --> 00:03:25.377 By introducing an orientation-based tracking approach. I've included some of our robust balancing results where our controller is designed to maintain the center of mass between the two While an operator applies forces to the pose. There's a QR code on the poster if you want to see a video. 00:03:26.253 --> 00:03:34.253 I've included the actuator tracking results for one of the legs, validating the joint torque control design that is robust to the unmodeled structural elasticity. 00:03:34.191 --> 00:03:50.191 You can also see a comparison between the physical robot and the virtual system so that you can directly understand some of the kinematic challenges that we've had to deal with. And in sum, the joint tour controller is able to maintain a 12 hertz bandwidth. And this dissertation has results that show a 50% 00:03:49.880 --> 00:03:54.880 Drop in joint state error and stability improvements using the proposed compensator. 00:03:54.628 --> 00:04:05.628 In conclusion, this work enables 3D printed compliant humanoids with robust balancing capabilities and provides results for a robot size that has not previously been shown. 00:04:05.691 --> 00:04:32.691 Only the humanoid's lower body was built because this project was never funded. So future work would focus on Including an upper body design that would also help the walking performance by raising the center of mass. More work would need to go into identifying dynamic modeling approaches for the structural elasticity, or this work avoided considering things like additional degrees of freedom, nonlinear stiffness and friction models. 00:04:32.752 --> 00:04:39.752 As well as closed chain dynamic models, specifically trying to avoid challenges with real-time constraints. 00:04:39.690 --> 00:04:52.690 In addition, the entire robot design has been included open source, including the robots CAD, electrical schematics, microcontroller firmware, and high-level software for the benefit of the research community. 00:04:53.128 --> 00:04:58.128 I've included my references as well as feature publications at the bottom, and I'd like to thank you for your time.