11 July 2018
|09.15-09.30||Bharadwaj Amrutur||Welcome Address|
|09.30-10.30||Raj Rajkumar||Self-driving vehicles: The road ahead
Self-driving vehicles have become very popular in the public mindset over just the past few years, triggered in good part by the DARPA Grand Challenges of the 2000’s. Self-driving vehicles indeed hold the potential to revolutionize the transportation landscape along multiple dimensions. This talk will highlight these transformative opportunities, while raising some basic questions that need to be addressed for the revolution to materialise. These issues include the technological barriers that currently prevent vehicles to become completely driverless, constraints on sensing and recognition, the role of connectivity, and cost considerations. Also, liability, insurance, regulations and societal acceptance will also impact adoption. The talk will be based on on-road experiences and will inject some speculation to trigger an active discussion.
|10.30-11.30||Tobias Altmüller||Autonomous vehicles as 3rd living space: Challenges driven by usability and user experience
Automated driving will bring new mobility concepts and allow the passengers to gain free time while being driven. The car as third living space converts driving time into quality of life. It gives the passengers the extra leeway to do the things they want to do.
This talk addresses the potential of the car as third living space and gives some use cases of what people would like to do in automated cars. There are different drivers which are increasing system as well as interaction complexity of such cars. Intelligent systems will help to handle this complexity. On the other hand there are additional challenges coming up for development of HMI with intelligent systems to achieve high user experience.
This talk will show a vision of the car as third living space, point out some of the challenges and how they are addressed.
|11.30-12.00||Break||Ballroom (next to Faculty Hall)|
|12.00-13.00||Suhas Diggavi||Security and privacy in cyber-physical systems
Many of our everyday activities are increasingly dependent on cyber-physical systems (CPS), which are systems capable of communication & computation and that interact with the physical world through sensing and actuation. The security of such systems is critical, since an adversary can cause damage in the real-world, not just compromise information as in today’s internet, as has been demonstrated by attacks on nuclear plants, smart cars, medical devices, electric grids, etc. To secure CPS systems, just protecting bits (cyber-security) is insufficient: A sensor attack can feed wrong inputs to sensors and thus manipulate the physical signals before they get converted to bits. Moreover, in order to learn behaviours from CPS and take decisions/actions, in many applications, one needs to also develop privacy guarantees, so that the collected data does not compromise sensitive information.
In this talk we will develop a framework for security and privacy of CPS by focusing on what makes these notions different from information security and privacy. We review work on secure state estimation and control for CPS that draw insights from real error correction. We also develop privacy framework for CPS that could enable light-weight CPS-specific privacy mechanisms. We will conclude with some open questions and challenges in CPS security and privacy.
|13.00-14.00||Lunch||Main Guest House|
|14.00-15.00||Rahul Jain||Reinforcement learning for control of unknown autonomous systems
System identification followed by control synthesis has long been the dominant paradigm for control engineers. And yet, many autonomous systems must learn and control at the same time. Adaptive Control Theory has indeed been motivated by this need. But it has focused on asymptotic stability while many contemporary applications demand finite time (non-asymptotic) performance optimality. Results in stochastic adaptive control theory are sparse with any such algorithms being impractical for real-world implementation.
I propose Reinforcement Learning algorithms inspired by recent developments in online (bandit) learning. Two settings will be considered: Markov decision processes (MDPs) and Linear stochastic systems. I will introduce a posterior-sampling based regret-minimization learning algorithm that optimally trades off exploration vs. exploitation and achieves order optimal regret. This is a practical algorithm that obviates the need for expensive computation and achieves non-asymptotic regret optimality. I will then talk about a general non-parametric stochastic system model on continuous state spaces. Designing universal control algorithms (that work for any problem) for such settings (even with known model) that are provably (approximately) optimal has long been a very challenging problem in both Stochastic Control and Reinforcement Learning. I will propose a simple algorithm that combines randomized function approximation in universal function approximation spaces with Empirical Q-Value Learning which is not only universal but also approximately optimal with high probability. In closing, I would like to argue that the many problems of Reinforcement Learning present a new opportunity for Controls and some interesting future directions.
|15.00-16.00||Carl-Gustaf Jansson||Smart City initiatives based on engineering of cyber-physical systems and application of Artificial Intelligence techniques
The purpose of this talk is to illustrate the power of the concept of cyber-physical systems (CPS) as a key component in several Smart City initiatives around the globe. CPS means the introduction of highly connected, complex and co-engineered systems of physical and computational components. CPS will play a major role in solutions to a large set of urban problems, e.g. traffic flow management, energy production & delivery, water and waste management, emergency responsiveness, logistics, manufacturing and personalized health care. The concept of CPS includes related but more limited concepts as Internet of Things and Embedded Systems. The theoretical basis for CPS is a combination of cybernetics and a broad range of theories from other fields. Artificial Intelligence share with Cybernetics a general loop of perception -> analysis/reflection -> adaption -> action. Such a loop is also typically the backbone of many Smart City engineering solutions. It is advocated that one of the major reason for the current renewed strong interest in Artificial Intelligence (AI) is that, since long established AI methods and algorithms suddenly get immensely more muscle power when applied in the CPS scenarios out-lined above. Other contributing factors are the immensely increased availability of data and the strong growth of computational resources.
|16.00-18.00||Poster and Demo Session||Ballroom (next to Faculty Hall)|
|18.00-20.00||Networking Dinner||Main Guest House|
12 July 2018
|09.30-10.30||Li Zhang||Magnetic swimming microrobots: From individual control to swarming behaviour
People have envisioned tiny machines and robots that can explore a human body, find and treat diseases since Richard Feynman’s famous speech, “There’s plenty of room at the bottom,” in which the idea of a “swallowable surgeon” was proposed in the 1950s. Even though we are at a state of infancy to achieve this vision, recent intense progress on nanotechnology, MEMS/NEMS technology and micro-/nanorobotics has accelerated the pace toward the goal. A number of research efforts have been recently published regarding the development of swimming tiny machines/robots from the basic principles and fabrication methods to practical applications. This talk will cover both fundamental and applied aspects of magnetic swimming microrobots, from individual control to their collective behavior, and perspective of using these small agents for biological and biomedical applications will be discussed.
|10.30-11.30||Ayusman Sen||Fantastic voyage: Designing self-powered nanobots
Self-powered nano and microscale moving systems are currently the subject of intense interest due in part to their potential applications in nanomachinery, nanoscale assembly, robotics, fluidics, and chemical/biochemical sensing. One of the more interesting recent discoveries has been the ability to design nano/microparticles, including molecules, which catalytically harness the chemical energy in their environment to move autonomously. These “bots” can be directed by chemical and light gradients. Further, our group has developed systems in which chemical secretions from the translating micro/nanomotors initiate long-range, collective interactions among the particles. This behavior is reminiscent of quorum sensing organisms that swarm in response to a minimum threshold concentration of a signaling chemical. In addition, an object that moves by generating a continuous surface force in a fluid can, in principle, be used to pump the fluid by the same catalytic mechanism. Thus, by immobilizing the nano/micromotors, we have developed nano/microfluidic pumps that transduce energy catalytically. These non- mechanical pumps provide precise control over flow rate without the aid of an external power source and are capable of turning on in response to specific analytes in solution. In addition, the catalytic pumps can be harnessed for directional delivery of microparticles in specific locations in space. We will discuss recent experimental results, as well as approaches to the modeling of the complex emergent behavior of these particles.
|11.30-12.00||Break||Ballroom (next to Faculty Hall)|
|13.00-14.00||Lunch||Main Guest House|
|14.00-15.00||Bhaskar Krishnamachari||Architectures and algorithms for the future Internet of Things
The Internet of Things (IoT) is starting to grow exponentially with applications ranging from consumer electronics and smart homes to intelligent transportation and smart cities. We present an overview of more than a decade of academic research on algorithms and protocols for IoT at the University of Southern California (USC), from networking low-power devices to location services, as well as ongoing research and development efforts pertaining to privacy-sensitive and secure middleware, experimental testbeds, and online reinforcement learning for IoT systems at the USC Viterbi Center for Cyber-Physical Systems and the Internet of Things. We also describe our ongoing work on developing a large-scale campus-wide IoT testbed at USC. And we highlight an interdisciplinary effort to develop and implement a new “marketecture” that we refer to as I3 (for Intelligent IoT Integrator). This work aims to provide incentives for device owners to share real time streaming sensor data as well as access to actuators with third-party application developers, and as an open ecosystem that allows for buyers, sellers and data-processing brokers to easily and scalably link to each other.
|15.00-16.00||Balaji Prabhakar||Self-programming networks: Architecture and algorithms
We describe self-programming networks (SPNs), an ongoing research effort at Stanford for making Cloud computing networks autonomous; that is, to enable networks to sense and monitor themselves, and program and control themselves. We present the architecture of SPNs and two key algorithms: (i) Simon, for fine-grained network measurement using packet and probe timestamps taken at the network’s edge, and (ii) Huygens, for nanosecond-level clock synchronisation in real-time and at scale. We will present the algorithms and results from some deployments.
|16.00-16.30||Networking Break||Ballroom (next to Faculty Hall)|