Schedule

8:30 Welcome & Introduction by the Organizers
8:40 Michael Hamilton
Drones, Nodes, and Apps: perspectives and prospects for the next generation of ecological applications using Micro Aerial Vehicles
9:15 Victor Hernandez Bennetts
Mobile Robotics Olfaction: Towards Practical Applications
9:35 Fabio Ramos
Beyond Information-Gain Exploration: Bayesian Optimisation for Smart Planning
10:00 Coffee Break
10:20 Poster Spotlights
10:50 Poster Session
11:50 Lunch/Soccer Break
15:00 Larry Matthies
Autonomous Aerial Mobility on Earth and Other Planets
15:35 Tim Barfoot
Towards Visual Navigation to Support Long-Term Robotic Monitoring
16:05 Paul Scerri
Monitoring Water: Data and Lessons from the Field
16:30 Coffee Break
17:00 Ryan Eustice
Robust and Persistent Visual SLAM for Autonomous Underwater Hull Inspection and Monitoring
17:25 Mike Bosse
Discrete to Continuous Trends at ETH ASL
17:45 Discussion & Concluding Remarks by the Organizers
18:45 End of workshop

Venue

  • Location: Wheeler Hall, room 101
  • Date: July 13, 2014
  • Registration

Invited Speakers

Larry Matthies, NASA JPL

Autonomous Aerial Mobility on Earth and Other Planets
Small, autonomous, aerial vehicles with vertical take-off and landing capability have a wide variety of potential applications on Earth and, as it turns out, in solar system exploration. The autonomous navigation capabilities needed for such systems also have other applications in solar exploration scenarios. This talk will give an overview of recent progress at JPL on autonomous navigation capabilities for micro air vehicles on Earth, covering vision-aided state estimation, obstacle avoidance, and roof-top landing, then survey potential applications for similar capabilities in solar system exploration for missions to Mars, Titan, Venus, comets, and asteroids.


Michael Hamilton, UC Berkeley

Drones, Nodes, and Apps: perspectives and prospects for the next generation of ecological applications using Micro Aerial Vehicles
Field stations have been centers of environmental science for a century — the places where scientists go to study environmental processes in their natural context. They are crucibles of innovation and discovery, storehouses of increasingly critical historical information, and hubs of integrated research; they have the infrastructure to support complex experiments and to maintain long-term projects. As investments in distributed wireless remote sensing systems at field stations mature in their benefits to our characterization and understanding of complex ecological and environmental processes across varied habitats and diverse landscapes, an emerging class of sensing platform, the micro aerial vehicle (MAV), promises to stimulate further opportunities for advancing field research and science education.
For several years, engineers visiting Blue Oak Ranch Reserve have been developing MAV applications to scan vegetation, delineate wildfire behavior, characterize the hydro-periods of vernal wetlands, and non-intrusively observe wildlife. Better lighter cameras, improved autonomous navigation, and longer flight times are enabling improvements in how we can cost-effectively and non-destructively document plant growth and phenology -- and a new class of MAV which can automatically and precisely pipette water from ponds and streams at predetermined depths and sampling intervals is now undergoing testing at the Reserve.
Before long, field ecologists and students will have in-situ access to geo-referenced environmental data streams, sourced from multiple ground and airborne sensors, delivered to their hand-held tablets with apps that integrate and fuse information with their own direct observations and measurements, to explore a new realm of ecological analysis and discovery at our field station.


Tim Barfoot, Toronto

Towards Visual Navigation to Support Long-Term Robotic Monitoring
At a fundamental level, the notion of monitoring has to do with gathering data over time in order to detect changes. Mobile robots can be used to traverse environments repeatedly, but this can present navigation challenges due to the very changes being monitored. In this talk, I will make the argument that long-term autonomy and change detection are two sides of the same coin; we need to be able to correlate old data with new to navigate, but also to recognize both acute and gradual changes. I will frame the discussion in the context of a visual route-repeating technique for ground robots operating in GPS-denied environments. I will discuss various features we have incorporated into the method in order to make it work long term (e.g., lighting invariance), as well as plans for the future.


Fabio Ramos, ACFR/University of Sydney

Beyond Information-Gain Exploration: Bayesian Optimisation for Smart Planning
Information gain approaches to exploration make a strong assumption that all parts of the environment are equally important, and the goal is to minimise the overall uncertainty. However, in many problems such as air pollution monitoring or terrain reconstruction, areas with high levels of pollution or more complicated terrain shapes require denser sampling. In this talk I will present a novel technique for learning spatial-temporal environment processes where the amount of sampling depends on the complexity and/or magnitude of the function being learnt. The method is based on Bayesian optimisation and is able to select paths that maximise the prediction performance for processes where tracking peaks is crucial (such as air pollution), trading exploration-exploitation in a principled statistical manner. I will show applications in air pollution monitoring, vibration modelling while navigating on uneven terrains, and lightening changes to illustrate the benefits of the approach. Finally I will show a commercial application of the technique where a web-based app was built for the Environmental Protection Agency in Australia, for real-time air pollution forecast in the Hunter Valley region, where coal mines, urban centres and vineyards need to coexist.


Ryan Eustice, University of Michigan

Robust and Persistent Visual SLAM for Autonomous Underwater Hull Inspection and Monitoring
The field of simultaneous localization and mapping (SLAM) has made tremendous progress in the last couple of decades, to the point where we have mature-enough methods and algorithms to explore applications on interesting scales both spatially and temporally. In this talk we discuss some of our current efforts in deploying large-scale, long-term SLAM systems in real-world field applications, and in particular, our current work in autonomous underwater ship hull inspection. We will discuss our developments in modeling the visual saliency of underwater imagery for pose-graph SLAM, how this saliency measure can be used within an active SLAM planning paradigm, and our development of generic linear constraints---a principled framework for pose-graph reduction, which is important for controlling multi-session SLAM graph complexity.

Paul Scerri, CMU

Monitoring Water: Data and Lessons from the Field
In this talk, I will report on some of the water data we have collected with autonomous robotic boats. We have collected a variety of data in a variety of environments and learned a lot about the challenges of actually getting good data. Some of those challenges include dealing with hysteresis in the sensor, unpredictable complexity and doing consistent calibration.


Victor Hernandez Bennetts, Örebro

Mobile Robotics Olfaction: Towards Practical Applications
Mobile Robotics Olfaction (MRO) is the line of research in computer science that addresses the task of integrating chemical sensing modalities in mobile robotics platforms. MRO prototypes have progressed from toy-like robots that reactively followed chemical plumes in the early 1990s, to functional proof of concept prototypes, able to carry out gas sensing in realistic environments. MRO systems are a promising technology that can be used in different civil, industrial and safety related applications. In this talk, we present an overview of MRO, where we identify the challenges and tasks that a fully fledged robotic system has to address to perform gas sensing in realistic environments. In addition, we present an overview of the current work being developed at the at the Mobile Robotics and Olfaction Lab in Örebro in Sweden, which is one of the lead research institutes in this field.


Mike Bosse, ETH Zurich

Discrete to Continuous Trends at ETH ASL

Contributed Posters