Monday, February 23, 2015

2015 SEAS Penn iTalks finalists

7:00-8:00 pm: SEAS Penn iTalks presentation at Wu and Chen Hall
8:00-9:30 pm: Happy hour (open to all graduate engineering students) at Levine Lobby
The three finalists presenting during this session are:

Session 1: March 25th

Adrian Lievano and Matthew Lisle (MEAM/Robotics)

BionUX, The New Era for Upper-Limb Prosthetic Devices

3D-PO will be presenting the most advanced version of a 3D-Printed prosthetic for upper-limb amputees. In addition to using EMG(muscle) controlled grasp techniques, we are pushing the boundaries of low-cost prosthetics and adding grip-force and vibrotactile sensory feedback to the user. We will present the prior art, the current market, an evaluation of low-cost and state of the art prosthetics, our current work, and our future vision and potential of this technology.

Madhur Behl (ESE)

Sometimes, Money Does Grow On Trees

The most recent decade was the nations warmest on record and experts predict that temperatures are only going to rise. Every year, heat waves in summer and polar vortexes in winter are growing longer in duration and cause massive peaks in the electricity power consumption across the electric grid. The peaks in the electricity consumption overburden, an already overstressed grid and can result in energy shortages and blackouts. For Penn's campus, which has several hundred buildings, the electricity bill for just 5 days is up to $1.5 million due to peak power pricing imposed by the utility companies. This talk will be about DR-Advisor: A Data Driven Demand Response Recommender System. Demand response (or DR) refers to intentional changes in electricity usage by end-use customers from their normal consumption patterns in response to changes in the price of electricity, or to financial payments provided by utilities during peak hours or when electric grid reliability is jeopardized.Using just historical power consumption and weather data, we build family of regression trees to learn models for predicting the real-time power consumption of a building. These trees are used for making recommendations for the building’s facilities manager on the best electricity curtailment strategies for demand response, which minimize electricity costs. The techniques and methods that I will describe in this talk are at the intersection of large scale systems research, machine learning, control theory and statistics. 

Justin Thomas (MEAM)

Extending Autonomous Capabilities of Micro Aerial Vehicles for Real World Applications

Micro Aerial Vehicles (MAVs) have the potential to improve many fields such as first response, environmental monitoring, and package delivery. However, MAVs are currently restricted to aerial observation roles and cannot physically interact with their surroundings. Further, with maximum flight times around thirty minutes, mission durations are significantly limited, and the use of MAVs is not yet practical in many applications. In observation and surveillance tasks, a robot could perch and shut off the motors to decrease energy usage, and in other tasks, we can decrease the required airborne time by increasing the speed of grasping. In this presentation, I will show preliminary steps to achieve autonomous, gecko-inspired perching and the fastest-known, avian-inspired grasping using quadrotors.

Session 2: April 1st

Heather Culbertson (MEAM)

Haptics: Making the Virtual World Feel Real

The haptic sensations felt when interacting with physical objects create a rich and varied impression of the objects and their environment. The human sense of touch excels at sensing and interpreting these haptic cues, even when the object is felt through an intermediary tool instead of directly with a bare finger. Dragging, pressing, and tapping a tool on the object allow you to sense the object's roughness, slipperiness, and hardness as a combination of vibrations and forces. Unfortunately, the richness of these interaction cues is missing from many virtual environments, leading to a less satisfying and immersive experience. However, we can create the perceptual illusion of touching a real object by displaying the appropriate haptic signals during virtual interactions. This talk presents my work in creating haptic models of textured surfaces from acceleration, force, and speed data recorded during physical interactions. I will discuss the two haptic rendering systems I have created for displaying virtual surfaces using these models. An evaluation of these systems has shown that we have created the most realistic haptic surfaces to date.

Amin Rahimian (ESE)

Learning without Recall

People exchange beliefs in social networks to benefit from each other's opinions and private information
in trying to learn an unknown state of the world. Beliefs about the unknown are mathematically modelled as probability distributions over the set of finitely many possibilities, and the refinement of beliefs with new observations is therefore understood as an update from one probability distribution to another. The rational (optimal) approach to this problem of social learning is for each agent to successively apply the Bayes' rule to her entire history of observations. However, it is well known that repeated applications of Bayes' rule in networks become computationally intractable, since the rational agent would need to make very complicated inferences about the possible sources of her observations. This motivates the "Learning without Recall" model of belief propagation, which we consider in this talk. In this model, although agents behave rationally, they do not recall their history of past observations. We analyze the evolution of beliefs amongst these so-called "Rational but Memoryless" agents and show that they can still learn the truth by relying on each other’s observations; provided that as time evolves they put less and less weight on their neighbouring beliefs - eventually learning the truth, but in isolation.

Sarah Tang (MEAM)

Planning for Aggressive Maneuvering of a Quadrotor with a Cable-Suspended Load

Unmanned aerial vehicles (UAVs), in particular quadrotors, have emerged as popular platforms for both research institutions and startup companies. We are interested in using quadrotors for autonomous payload transportation. Past research in this area has focused solely on stabilization of the load and minimization of the load swing while traversing trajectories, severely limiting the system's range of motion. This prohibits the use of autonomous quadrotors for many tasks completed by skilled helicopter pilots today, such as Christmas tree harvesting and the extinguishing of large forest fires. Unlike previous works, our models allow for complex interactions with the load, such as load pick-ups and releases and even small periods of zero cable tension where the load temporarily undergoes free-fall before being “caught" again by the quadrotor. My talk will discuss our techniques for designing these types of maneuvers to navigate a quadrotor with a cable-suspended payload through known obstacle-filled environments

Come cheer on your peers and vote for your favorite talk (the audience favorite prize is worth $100 and will be determined at the end of the session). After the talks, stay for the happy hour and network with the judging panel (engineering professors), as well as the presenters and fellow engineering graduate students. The happy hour portion of this event is for students who are 21+. 

If you have any questions/feedback about the event, please email