Special Symposium on Intelligent Systems 2018

Agenda

Tuesday, September 18, 2018, Stuttgart

10:00 - 10:30

10:30 - 11:15

It is exciting to imagine a future when we can use wearable robots to increase strength or efficiency, restore or repair ability after injury or prevent injuries from happening in the first place. This vision is currently challenging to achieve due to limitations in current technology and a lack of understanding of how humans will respond to physical assistance. This talk will give an overview of our work on developing disruptive soft wearable robot technologies for augmenting and restoring human performance and what we have learned from biomechanical and physiological studies that furthers the scientific understanding of how humans interact with such machines. Our efforts are the result of a multidisciplinary team of students and research staff with backgrounds in engineering, materials science, apparel design, industrial design, biomechanics, and physical therapy, in addition to valuable collaborations with colleagues from Harvard, Boston University, and beyond. Our long-term vision is for ubiquitous soft wearable robots that can be worn all day, every day underneath clothing, in both the community, home, sporting and workplace environments. This would enable a change in the paradigm of how we currently assist mobility for those with physical impairments (i.e. replacing plastic braces with active assistance), and how we rehabilitate those after injury (i.e. training programs outside of a clinic environment) as well as how we protect healthy individuals with physical tasks (i.e. protecting military personnel, athletes or factory workers performing high risk tasks).

Speaker Bio >>

11:15 - 11:45

Coffee Break (In front of Lecture Hall 2R04, Stuttgart)

11:45 - 12:30

Many organisms fly in order to survive and reproduce. My lab focusses on understanding bird flight to improve flying robots—because birds fly further, longer, and more reliable in complex visual and wind environments. I use this multidisciplinary lens that integrates biomechanics, aerodynamics, and robotics to advance our understanding of the evolution of flight more generally across birds, bats, insects, and autorotating seeds. The development of flying organisms as an individual and their evolution as a species are shaped by the physical interaction between organism and surrounding air. The organism’s architecture is tuned for propelling itself and controlling its motion. Flying animals and plants maximize performance by generating and manipulating vortices. These vortices are created close to the body as it is driven by the action of muscles or gravity, then are ‘shed’ to form a wake (a trackway left behind in the fluid). I study how the organism’s architecture is tuned to utilize these and other aeromechanical principles to compare the function of bird wings to that of bat, insect, and maple seed wings. The experimental approaches range from making robotic models to training birds to fly in a custom-designed wind tunnel as well as in visual flight arena’s—and inventing methods to 3D scan birds and measure the aerodynamic force they generate—nonintrusively—with a novel aerodynamic force platform. The studies reveal that animals and plants have converged upon the same solution for generating high lift: A strong vortex that runs parallel to the leading edge of the wing, which it sucks upward. Why this vortex remains stably attached to flapping animal and spinning plant wings is elucidated and linked to kinematics and wing morphology. While wing morphology is quite rigid in insects and maple seeds, it is extremely fluid in birds. I will show how such ‘wing morphing’ significantly expands the performance envelope of birds during flight, and will dissect the mechanisms that enable birds to morph better than any aircraft can. Finally, I will show how these findings have inspired my students to design new flapping and morphing aerial robots.

Speaker Bio >>

13:45 - 14:30

Soft robotics deals with interaction with environments that are uncertain and vulnerable to change, by easily adapting to the environment with soft materials. However, softness requires controlling large degrees of freedom. Many soft robots use pneumatics which can easily distribute the actuation. If tendons are used for actuating a soft body, the large degrees of freedom of the material either requires large number of tendons or limits the controllability. Tendon drive soft robots can benefit from using the concept of physically embodied intelligence, first proposed by Prof. Rolf Pfeifer. By embodying intelligence into the design, better performance can be achieved with a simpler actuation. In nature, there are few example that exhibit this property. Flytrap, for example, can close its leaves quickly by using bistability of the leaves instead of just relying on the actuation. Inchworm achieves adaptive gripping with its prolegs by using the buckling effect. In this talk, I will give an overview of various soft b robotic technologies, and some of the soft robots with physically embodied intelligence that are being developed at SNU. These examples will show that the concept of physically embodied intelligence simplifies the design and enables better performance by exploiting the characteristics of the material.

Speaker Bio >>

14:30 - 15:15

Robots are moving from the factory floor and into our lives (e.g. autonomous cars, package delivery drones, and search-and-rescue devices). However, compared to living systems, locomotion by such devices is still relatively limited, in part because principles of interaction with complex environments are largely unknown. In this talk I will discuss efforts to develop a physics of moving systems -- a locomotion ``robophysics'' -- which we define as the pursuit of the discovery of principles of self-generated motion [Aguilar et al, Rep. Prog. Physics, 2016]. We use the methods of physics to examine successful and failed locomotion in simplified laboratory devices using parameter space exploration, systematic control, and techniques from dynamical systems. Drawing from examples from my group and our collaborators, I will discuss how robophysical studies in terrestrial environments have begun to aid engineers in the creation of devices that begin to achieve life-like locomotor abilities on and within complex environments, have inspired new physics questions in low-dimensional dynamical systems, geometric mechanics and soft matter physics, and have been useful to develop models for biological locomotion in complex terrain. The rapidly decreasing cost of constructing sophisticated robot models with easy access to significant computational power bodes well for scientists and engineers to engage in a discipline which can readily integrate experiment, theory and computation.

Speaker Bio >>

Wednesday, September 19, 2018, Stuttgart/Tübingen

10:00 - 10:45

The ultimate goal of any soft robotics system is to have a cohesive solution to improve the human – machine interface. For such an interface, it is critical to realize a versatile and adaptable multi-degrees of freedom robotic design.  While the findings in soft robotics have broadened the application of robotics, they are still limited to specific scenarios. The next challenge is in pushing the boundaries of multi-disciplinary science interceptions simultaneously: materials, mechatronics, energy, control, and design. Such efforts will lead to robust solutions in design methodology, novel actuators, and a comprehensive fabrication and integration method of the core robotic components. This talk will highlight on the recent progresses in soft- material robots and origami robots that aim at achieving comprehensive solutions toward diverse soft human – robot applications. 

Speaker Bio >>

10:45 - 11:30

Robots generally excel at specific tasks in structured environments, but lack the versatility and adaptability required to interact-with and locomote-within the natural world. To increase versatility in robot design, my research group is developing robotic skins that can wrap around arbitrary deformable objects to induce the desired motions and deformations. Robotic skins integrate actuation and sensing into a single conformable material, and may be applied-to, removed-from, and transferred-between different objects to create a multitude of controllable robots with different functions to accommodate the demands of different environments. We have shown that attaching the same robotic skin to a deformable object in different ways, or to different objects, leads to unique motions. Further, we have shown that combining multiple robotic skins enables complex motions and functions. During this talk, I will demonstrate the versatility of this soft robot design approach by showing robotic skins in a wide range of applications - including manipulation tasks, locomotion, and wearables - using the same 2D robotic skins reconfigured on the surface of various 3D soft, inanimate objects.

Speaker Bio >>

11:30 - 12:00

Coffee Break (In front of Lecture Hall 2R04 Lecture Hall 2R04, Stuttgart)

12:00 - 12:45

Robotics has the potential to address many of today’s pressing problems in fields ranging from healthcare to manufacturing to disaster relief. However, the traditional approaches used on the factory floor do not perform well in unstructured environments. The key to solving many of these challenges is to explore new, non-traditional designs. Fortunately, nature surrounds us with examples of novel ways to navigate and interact with the real world. Dr. Tolley’s Bioinspired Robotics and Design Lab seeks to borrow the key principles of operation from biological systems and apply them to robotic design. This talk will give an overview of projects in the lab demonstrating approaches to the design, fabrication, and control of soft robotic systems. These projects seek to develop bioinspired systems capable of navigating the world by walking, digging, and swimming, of interacting directly with humans and delicate objects, and of self-assembly by folding.

Speaker Bio >>

14:00 - 14:45

Robots today rely on rigid components and electric motors based on metal and magnets, making them heavy, unsafe near humans, expensive and ill-suited for unpredictable environments. Nature, in contrast, makes extensive use of soft materials and has produced organisms that drastically outperform robots in terms of agility, dexterity, and adaptability. The Keplinger Lab aims to fundamentally challenge current limitations of robotic hardware, using an interdisciplinary approach that synergizes concepts from soft matter physics and chemistry with advanced engineering technologies to introduce intelligent materials systems for a new generation of life-like robots. One major theme of research is the development of new classes of actuators – a key component of all robotic systems – that replicate the sweeping success of biological muscle, a masterpiece of evolution featuring astonishing all-around actuation performance, the ability to self-heal after damage, and seamless integration with sensing.

This talk is focused on the labs' recently introduced HASEL artificial muscle technology. Hydraulically Amplified Self-healing ELectrostatic (HASEL) transducers are a new class of self-sensing, high-performance muscle-mimetic actuators, which are electrically driven and harness a mechanism that couples electrostatic and hydraulic forces to achieve a wide variety of actuation modes. Current designs of HASEL are capable of exceeding actuation stress of 0.3 MPa, linear strain of 100%, specific power of 800W/kg, full-cycle electromechanical efficiency of 30% and bandwidth of 100Hz; all these metrics match or exceed the capabilities of biological muscle. Additionally, HASEL actuators can repeatedly and autonomously self-heal after electric breakdown, thereby enabling robust performance for millions of cycles. Modeling results predict the impact of material parameters and scaling laws of these actuators, laying out a roadmap towards future HASEL actuators with drastically improved performance. These results highlight opportunities to further develop HASEL artificial muscles for wide use in next-generation robots that replicate the vast capabilities of biological systems.

Speaker Bio >>

15:45 - 16:00

Coffee Break (Lounge, Tübingen)

16:00 - 16:30

Thursday, September 20, 2018, Tübingen

10:15 - 11:00

Being able to autonomously learn from scratch is a key ability of intelligent systems - and a central focus of my research. A particular challenge in real world robotic scenarios are methods that are in addition highly data-efficient and robust, since data-collection on real robots is time intensive and often expensive. I will discuss two main areas of progress towards this goal - improved off-policy learning and better exploration - and give examples of simulated and real robots that can learn increasingly complex tasks from scratch.

Speaker Bio >>

11:00 - 11:45

The ultimate promise of robotics is to design devices that can physically interact with the world. To date, robots have been primarily deployed in highly structured and predictable environments. However, we envision the next generation of robots (ranging from self-driving and -flying vehicles to robot assistants) to operate in unpredictable and generally unknown environments alongside humans. This challenges current robot algorithms, which have been largely based on a-priori knowledge about the system and its environment. While research has shown that robots are able to learn new skills from experience and adapt to unknown situations, these results have been limited to learning single tasks, and demonstrated in simulation or lab settings. The next challenge is to enable robot learning in real-world application scenarios. This will require versatile, data-efficient and online learning algorithms that guarantee safety when placed in a closed-loop system architecture. It will also require to answer the fundamental question of how to design learning architectures for dynamic and interactive agents. This talk will highlight our recent progress in combining learning methods with formal results from control theory. By combining models with data, our algorithms achieve adaptation to changing conditions during long-term operation, data-efficient multi-robot, multi-task transfer learning, and safe reinforcement learning. We demonstrate our algorithms in vision-based off-road driving and drone flight experiments, as well as on mobile manipulators.

Speaker Bio >>

11:45 - 12:15

Coffee Break (Lounge, Tübingen)

12:15 - 13:00

In recent years, deep neural networks have become an ubiquitous tool in a broad range of AI applications, and inspired new models for neural processing in the brain. I will talk about similarities and discrepancies of how decisions are formed in biological and artificial neural networks and how this relates to the ability of domain adaptation, few-shot learning, task transfer and adversarial robustness in the context of visual perception. I will argue that a deeper understanding of the distributed nature of neural decision making will be crucial for more data-efficient, interpretable and robust learning machines and conclude with more speculative ideas on developing social intelligence in artificial neural networks.

Speaker Bio >>

14:15 - 15:00

Intelligent agents need to understand the world and to plan their actions at multiple levels of temporal resolution. In reinforcement learning, the value function is a standard way to model knowledge related to long-term action outcomes. In this talk, I will describe generalized value functions, which are a way of representing knowledge about many behaviours and at many temporal scales. I will describe how such a representation can be learned form experience and describe new open research questions in this area.

Speaker Bio >>

15:00 - 15:30

Coffee Break (Lounge, Tübingen)

15:30 - 16:15

Implicit generative modeling has recently scratched the surface on how deep learning can be used as a universal simulator. Until recently, deep learning has been used quite successfully to solve long standing discriminative problems in computer vision, speech and natural language processing, basically showing that hand-coded human-engineered features are suboptimal in the presence of: complex problems in which human only have a basic understanding of the variability of the data; and, the availability of large labelled data sets. Recently, Variational Auto-Encoders and Generative Adversarial Networks have shown that the same representation learning can be used for generative modelling. These implicit generative models do not provide an interpretable model for the available data, but a universal simulator that it is able to generate data similar to the one used for training. These tools can be used to simplify complex simulations (e.g. climate models) or limited observations (e.g. cosmology or particle physics), opening the door to Artificial Intelligence powered advances in many different fields of science. In this talk, we will first present these general methods and what are their potential used and current short comings. In the second part of the talk, we will focus on a recent application of GANs for password guessing. This is an ideal application to understand the need for GANs and understand why they work and what are their limitations. 

Speaker Bio >>