| 10:00 | Shipeng Liu, Cristina G. Wilson, Bhaskar Krishnamachari, and Feifei Qian | Title: Human Geoscientist Objective Functions for Robot-aided Field Data Collection Decisions
Abstract: Geoscientists can use field robots to explore ongoing global changes like desertification, the process whereby fertile lands change to desert. High spatiotemporal resolution data collected by robots can inform understanding of the complex causes of desertification, critical to preventing additional land loss and protecting biodiversity. A key challenge in developing more intelligent robots, that move beyond mobile data collection devices and begin to aid human experts with sampling decisions, is the lack of understanding of how scientists make such decisions and adapt strategies when presented with new information. In this study we examined the dynamic data collection decisions of 108 expert geoscientists using a simulated field scenario. Human data collection behaviors suggested two distinct objectives: an objective to improve information coverage, and an objective to verify beliefs about the hypothesis. We developed a highly-simplified quantitative decision model that allows the robot to predict potential human data collection locations based on the two observed human data collection objectives. Predictions from the simple model successfully captured sampling location choices for 70% of human expert scientists, and revealed a transition in objectives as the level of information increased. The findings will enable decision support algorithms that allow robotic teammates to infer experts’ desired data collection strategy based on abstract scientific objectives, in the long-term supporting the development of cognitively-compatible robotic field assistants. |
| 10:03 | Alice K. Li, Yue Mao, Sandeep Manjanna, Sixuan Liu, Victoria M. Edwards, Fernando Cladera Ojeda, Michael Anoruo, Jasleen Dhanoa, Bharg Mehta, M. Ani Hsieh, Douglas J. Jerolmack and Hugo N. Ulloa | Title: Towards Understanding Underwater Weather Events in Rivers Using Autonomous Surface Vehicles
Abstract: Climate change has increased the frequency and severity of extreme weather events such as hurricanes and winter storms. The complex interplay of floods with tides, runoff, and sediment creates additional hazards — including erosion and the undermining of urban infrastructure — consequently impacting the health of our rivers and altering ecosystem health. Observations of these underwater phenomena are rare, because satellites and sensors mounted on aerial vehicles cannot penetrate the murky waters. Autonomous Surface Vehicles (ASVs) provides a means to track and map these complex and dynamic underwater phenomena. This work highlights preliminary results and plans for high-resolution data gathering with ASVs along the lower Schuylkill River in the Philadelphia area. The preliminary results presented will guide future surveys where ASVs, equipped a suite of water quality sensors, will be deployed to measure and characterize various physical and chemical properties of the river during various weather conditions. These datasets will assess: changes in bathymetry following floods; the relation of mixing and stagnation zones to water quality; salt pollution from roads; and resuspension of fine sediment. In the future, adaptive sampling strategies will be used to strategically and autonomously sample from regions of maximum information content. |
| 10:06 | Monika Roznere, Mingi Jeong, Kizito Masaba, Jessica Volan Trough-Haney, David Lutz, Kathryn Cottingham, Michael Palace and Alberto Quattrini Li | Title: Towards Context-Based Sampling for Environmental Monitoring by Heterogeneous Robots and Remote Sensing Technologies
Abstract: This work discusses a sampling framework that exploits all data streams of a heterogeneous team of ASV, UAV, and satellites to best monitor an aquatic environment over long periods and to tackle open challenges for roboticists and limnologists. Exploitation in our case is context-based — on a given sampling day, not all data streams, sensors, or robots may be available, and thus our framework adapts the sampling team coordination to best compensate the missing or unreliable data streams. Based on our previous work on heterogeneous robot team sampling, and unlike most literature, this extended abstract will address the different concerns and ideas that arise when considering context-based sampling with more than average number or missing data streams. The context-based sampling framework and insights discussed in this paper are the foundation for the full implementation of an autonomous adaptive sampling system, which will be vital for long-term water quality monitoring to help inform water management and stakeholders on the impacts of climate change on local water bodies. |
| 10:09 | Zhiang Chen, Jnaneshwar Das and Melissa Wagner | Title: High-resolution tornado damage estimation using UAV and machine learning
Abstract: Tornado damage estimation is important for providing insights into tornado studies and assisting rapid disaster response. However, it is challenging to precisely estimate tornado damage because of the large volumes of perishable data. This study presents data-driven approaches to tornado damage estimation using imagery collected from Unpiloted Aerial Systems (UASs) following the 26 June 2018 Eureka Kansas tornado. High-resolution orthomosaics were generated from Structure from Motion (SfM). We applied deep neural networks (DNNs) on the orthomosaics to estimate tornado damage and assessed their performance in four scenarios: (1) object detection with binary categories, (2) object detection with multiple categories, (3) image classification with binary categories, and (4) image classification with multiple categories. Additionally, two types of tornado damage heatmaps were generated. By directly stitching the resulting image tiles from the DNN inference, we produced the first type of tornado damage heatmaps where damage estimates are accurately georeferenced. We also presented a Gaussian process (GP) regression model to build the second type of tornado damage heatmap (a spatially continuous tornado damage heatmap) by merging the first type of object detection and image classification heatmaps. The GP regression results were assessed with ground-truth annotations and National Weather Service (NWS) ground surveys. This detailed information can help NWS Weather Forecast Offices and emergency managers with their damage assessments and better inform disaster response and recovery. |
| 10:12 | Serena Mou, Dorian Tsai and Matthew Dunbabin | Title: Reconfigurable Robots for Scaling Reef Restoration
Abstract: Coral reefs are under increasing threat from the impacts of climate change. Whilst current restoration approaches are effective, they require significant human involvement and equipment, and have limited deployment scale. Harvesting wild coral spawn from mass spawning events, rearing them to the larval stage and releasing the larvae onto degraded reefs is an emerging solution for reef restoration known as coral reseeding. This paper presents a reconfigurable autonomous surface vehicle system that can eliminate risky diving, cover greater areas with coral larvae, has a sensory suite for additional data measurement, and requires minimal non-technical expert training. A key feature is an on-board real-time benthic substrate classification model that predicts when to release larvae to increase settlement rate and ultimately, survivability. The presented robot design is reconfigurable, light weight, scalable, and easy to transport. Results from restoration deployments at Lizard Island demonstrate improved coral larvae release onto appropriate coral substrate, while also achieving 21.8 times more area coverage compared to manual methods. |
| | |
| 14:45 | Vindula Jayawardana and Cathy Wu | Title: Reinforcement Learning for Eco-Lagrangian Control at Intersections
Abstract: Signalized intersections in arterial roads result in persistent vehicle idling and excess accelerations, contributing to fuel consumption and CO2 emissions. There has thus been a line of work studying eco-driving control strategies to reduce fuel consumption and emission levels at intersections. However, methods to devise effective control strategies across a variety of traffic settings remain elusive. In this paper, we propose a reinforcement learning (RL) approach to learn effective eco-driving control strategies. We analyze the potential impact of a learned strategy on fuel consumption, CO2 emission, and travel time and compare with naturalistic driving and model-based baselines. We further demonstrate the generalizability of the learned policies under mixed traffic scenarios. Simulation results indicate that scenarios with 100% penetration of connected autonomous vehicles (CAV) may yield as high as 18% reduction in fuel consumption and 25% reduction in CO2 emission levels while even improving travel speed by 20%. Furthermore, results indicate that even 25% CAV penetration can bring at least 50% of the total fuel and emission reduction benefits. |
| 14:48 | Avideh Zakhor, Emily Lathrop, Adam Curtis, Farbod Farshidian, Adit Roychowdry, Wai Naing, Jason Zang, Matthew Tang, Varun Saran and Ashwin Vangipuram | Title: Robotic Cleaning and Air Sealing of Attics in Single Family Residential Homes
Abstract: Buildings contribute to about one third of carbon emission in the U.S. EPA estimates that homeowners can save an average of 15% on heating and cooling costs by air sealing their homes and adding insulation in attics and floors over crawl spaces. Yet, today many U.S. attics are not being air sealed due to inaccessibility. There are three parts to attic retrofit: cleaning, air sealing, and blowing in new insulation. In this paper, we describe two robotic systems for cleaning and air sealing single family residential homes. We use off the shelf base robots with custom designed payload for each task. We have tested the vacuum cleaning robot in an actual home in Sonoma, CA. |
| 14:51 | Ibrahim Salman, Jason O’Kane and Ioannis Rekleitis | Title: Uniform coverage of large water bodies with islands under limited resources
Abstract: This paper addresses the problem of covering large bodies of water to collect quantitative measurements of water quality with an autonomous surface vehicle (ASV) to aid in monitoring and predicting harmful cyanobacteria blooms. The algorithm produces ASV trajectories that visit all representative areas of the environment, including circumnavigating islands, when present, by utilizing the skeleton of the skeleton method. This method finds the medial axis of the body of water, prunes redundant branches, then finds the midpoints (skeleton) between the trimmed medial axis and the boundaries of the area of interest. This work addresses the issue of separate components of produced trajectory in the presence of islands, by identifying individual trajectories as independent contours, then generating the shortest paths between these contours resulting in a continuous cyclic path covering the entire selected body of water in a single pass. |
| 14:54 | Soumya Sudhakar, Sertac Karaman and Vivienne Sze | Title: Carbon Emissions from Computing Onboard Autonomous Vehicles
Abstract: While there is great interest in applications where robotics can help reduce greenhouse gas emissions, less attention has been paid to the potential emissions from emerging applications in the field of robotics itself, such as autonomous vehicles (AVs). In this work, we analyze the scenario where AVs are widely adopted and show that the carbon emissions of computing onboard AVs have the potential to make a non-negligible impact on total emissions, comparable to the emissions from all data centers in 2018. We model the impact of computing onboard a global fleet of AVs on emissions by modeling the average computer power needed to process the autonomy stack, the average time driven, and the average carbon intensity of the electric grid for each AV. In order to capture the uncertainty associated with each of these variables in the projection of emissions and ground the projections in available data, we probabilistically model each variable, estimate parameters from available data, benchmark autonomy stacks on an Nvidia RTX 2080 Ti GPU, and simulate autonomy stacks on an Nvidia DRIVE Orin platform. Our analysis shows that in a widespread AV adoption scenario where approximately 95% of all vehicles are autonomous, a global fleet of AVs operating autonomy stacks that consume on average more than 1220 Watts results in emissions that equal or exceed emissions from all data centers in 2018 in more than 90% of modeled scenarios. While widespread adoption of AVs is not yet a reality, our work hopes to draw attention to this potential problem before it becomes another source of greenhouse gas emissions and encourage the design of software and hardware solutions that reduce the potential emissions.
|
| 14:57 | Bir Bikram Dey, Robert Codd-Downey, Michael Jenkin, James Zacher, Eva Blainey | Title: Environmental Monitoring Using a Semi Autonomous Surface Vehicle
Abstract: Invasive aquatic plant species, and in particular Eurasian Water-Milfoil (EWM), pose a major threat to domestic flora and fauna and can in turn negatively impact local economies. Numerous strategies have been developed to harvest and remove these plant species from the environment. However it is still an open question as to which method is best suited to removing a particular invasive species and the impact of different lake conditions on the choice. One problem common to all harvesting methods is the need to assess the location and degree of infestation on an ongoing manner. This is a difficult and error prone problem given that the plants grow underwater and significant infestation at depth may not be visible at the surface. Here we detail efforts to monitor EWM infestation and evaluate harvesting methods using an autonomous surface vehicle (ASV). This novel ASV is based around a mono-hull design with two outriggers. Powered by a differential pair of underwater thrusters, the ASV is outfitted with RTK GPS for position estimation and a set of submerged environmental sensors that are used to capture imagery and depth information including the presence of material suspended in the water column. The ASV is capable of both autonomous and tele-operation. |
| 15:00 | Leonard Bauersfeld and Davide Scaramuzza | Title: A Pen-and-Paper Method for Multicopter Speed and Range
Abstract: Multicopters are among the most versatile mobile robots with applications ranging from inspection and mapping tasks to monitoring large areas for wildfires or estimating the number of trees in a certain area. As such, they are a valuable asset to our collective effort to better document and understand climate change. In those settings, the range, endurance, and optimal speed a multirotor vehicle can achieve while performing its task is a decisive factor for vehicle design and mission planning.
We present an accurate pen-and-paper algorithm to estimate the range, endurance and optimal speed of multicopters to help future researchers build drones with maximal range and endurance, ensuring that future multirotor vehicles are even more versatile. When applied to numerous commercially available vehicles, the proposed method achieves an error below 10% in estimating the range, endurance and optimal speed of the multicopters. This advance is made possible by distilling the knowledge from a highly accurate and complex simulation into very few equations that cover the use-case common in aerial monitoring: straight-line flight at a constant speed. |
| 15:03 | Nils Wilde, Armin Sadeghi and Stephen Smith | Title: Learning Submodular Objectives for Team Environmental Monitoring
Abstract: We study the well-known team orienteering prob- lem where a fleet of robots collects rewards by visiting locations. Usually, the rewards are assumed to be known to the robots; however, in applications such as environmental monitoring and sampling the rewards are often subjective and specifying them is challenging. We propose a framework to learn the unknown preferences of the user by presenting alternative solutions to them, and the user provides a ranking on the proposed alternative solutions. We propose a framework to minimize a bound on the maximum deviation from the optimal solution, namely regret. Finally, we demonstrate the importance of learning user preferences in an extensive set of experimental results using real world datasets for environmental monitoring. |
| 15:06 | Kleio Baxevani, Grant Otto, Arthur Trembanis and Herbert Tanner | Title: Mission Planning and Control for Autonomous Surface Vehicles Conducting Environmental Surveys
Abstract: Environmental surveys are increasingly conducted using \acp{ASV} to increase safety, decrease carbon footprint, and provide force multiplication. This work reports on methodological and algorithmic modifications introduced into (a) an open source ROS-based mission planner for \acp{ASV} that extends its functionality with rapid in situ survey area capability, and (b) an \ac{ASV} path following controller in order to improve the quality of bathymetry data obtained. Simulation and experimental results are provided, demonstrating the new capability and confirming the anticipated performance improvements. |
| 15:09 | Joshua Ashley, Biyun Xie and Michael Sama | Title: Designing and Applying Sensor-Driven Flight Systems for Multi-UAV Environment Exploration
Abstract: In the event of a major airborne contaminant into the environment (e.g., gas leak), evacuation mechanisms would benefit from real-time or sensor-driven analysis to inform these mechanisms. Unmanned aerial vehicles (UAVs) provide a mobile sensing system that can better track contaminant distribution. This paper approaches the problem of trajectory planning in these scenarios as a robotic exploration problem and implements an algorithm for solving this model. The offline reinforcement learning technique soft actor-critic (SAC) is used to learn a policy that produces trajectories that maximize information gain. This algorithm was tested against a ‘greedy’ policy that only maximizes immediate reward. The algorithm was shown to produce significantly higher-quality trajectories on average. A system to implement the resulting policy in the real world is also discussed. |
| 15:12 | Madison S. Pickett, Hannah Kolano, Joseph Davidson and Aaron Marburg | Title: Design of a Robotic Testbed for Underwater Manipulation Research
Abstract: Remotely Operated Vehicles, or ROVs, are becoming an increasingly common research platform. However, testing new algorithms for them is a difficult process: simulations cannot properly capture the complexities of the underwater dynamics, but deploying an ROV in a pool or in sea trials is a costly, time-intensive, and risky operation. Therefore, the domain suffers from a lack of a middle-ground platform to test algorithms in realistic underwater conditions. To address this problem, we designed and fabricated an underwater test platform to emulate ROV operation in underwater conditions. |
