Title: Decentralised Multi-Robot Active Perception in Outdoor Environments
Abstract: Online decision-making in teams of autonomous field robots is a research area that has grown, from its roots in isolated information gathering tasks, to encompass rich, task-level problems that close the loop around planning and perception. These problems have a wide range of exciting applications in ground, aerial, and marine environments. In this talk, I will give an overview of recent results from our group in the theory and real-world application of multi-robot active perception using a variety of robot platforms. I will focus on the scholarly, environmental, and economic impact that this avenue of work has the potential to achieve, and that has already been realised through start-up companies or other commercial activity.
Title: Challenges in Deploying Robust Autonomy for Robotic Exploration in Marine Environments
Abstract: This talk will describe insights gained from a decade of autonomous marine systems development at the University of Sydney’s Australian Centre for Field Robotics. Over the course of this time, we have developed and deployed numerous underwater vehicles and imaging platforms in support of applications in engineering science, marine ecology, archaeology and geoscience. We have operated an Australia-wide benthic observing program designed to deliver precisely navigated, repeat imagery of the seafloor. This initiative makes extensive use of Autonomous Underwater Vehicles (AUVs) to collect high-resolution stereo imagery, multibeam sonar and water column measurements on an annual or semi-annual basis at sites around Australia, spanning the full latitudinal range of the continent from tropical reefs in the north to temperate regions in the south. The program has been very successful over the past decade, collecting millions of images of the seafloor around Australia and making these available to the scientific community through online data portals developed by the facility and affiliated groups. These observations are providing important insights into the dynamics of key ecological sites and their responses to changes in oceanographic conditions through time. We have also contributed to expeditions to document coral bleaching, cyclone recovery, submerged neolithic settlement sites, ancient shipwrecks, methane seeps and deepwater hydrothermal vents. The talk will also consider some of our more recent work focused on developing automated tools for working with this imagery and illustrate how this is being used to inform further exploration work using these platforms.
Title: Diving into the unknown: robotics as tools to study the deep-sea
Abstract: To know what is hidden underneath the ocean surface, humans have started investigating the ocean using indirect observation and survey methods from ships. However, to really understand many of the ocean processes, we need to get inside it. Technology has been playing a pivotal role in the exploration and understanding of the deep-sea. This is a place impossible to reach without technological aids due to the inhospitable environment, in particular the high pressure. These conditions make the direct observations of the seafloor expensive and technically more challenging than direct observations of the moon or mars surface (in fact, until today, we have observed directly only 5% of the Earth seafloor). The last 40 years were exceptionally rich in new discoveries at the deep-seafloor in large part due to the development of technologies and robots for sampling and for direct and indirect observations. One of these discoveries was made at the end of the last century when the first hydrothermal field was directly observed in the Pacific Ocean. As a deep-ocean scientist, this presentation will be focused on the role that marine research robots play in the sampling and direct observation of the seafloor, sharing also some experiences during oceanographic missions and giving examples of recent discoveries. Direct observations of the deep-ocean are carried out by underwater robots, which can now reach its deepest point (11,000 m). They can be classified as deep-submergence vehicles (DSV; manned robot which transport scientists to the bottom of the sea), remotely operated vehicles (ROV; unmanned robots controlled by scientists via cable on board of the ships) and autonomous underwater vehicles (AUV; a robot that navigates vast distances, collecting data without any human control). In addition, the overexploitation of ore resources on land has been increasing interest in deep-sea mineral resources, driving the development of gigantic robots that will be able to exploit resources from the seafloor in the near future.
Title: Monitoring Glaciers Beyond the Horizon
Abstract: Monitoring the dynamics of glaciers and ice sheets is a challenging, but necessary task to develop physical models which can assess the future of freshwater volumes in today’s warming climate, and the resulting sea-level rise. Unmanned aerial vehicles (UAVs) can be used to explore dangerous glacial environments relatively inexpensively while eliminating risks to humans. However, flight operation within remote, challenging terrain and potentially hazardous conditions as well as any desired physical interaction with the glacial terrain requires detailed on-board knowledge of the UAV’s environment as well as robust algorithms to locally handle any unexpected disturbances. This talk will present some ongoing research taking place at the Autonomous Systems Lab of ETH Zurich towards addressing challenges of operation within glacial regions. Results will further be presented from a BVLOS glacier monitoring campaign in Northwest Greenland using a solar-powered, fixed-wing UAV and a fully autonomous in situ sensor placement mission on a highly-crevassed glacier in the Swiss Alps using a multi-copter platform.
Title: Exploring the use of simple robots with informative path planning for complex outdoor environment
Abstract: Robots have been increasingly used to automate data collections for the understanding of environmental processes taking place in open oceans and large water bodies. However, their applications in urbanised coastal waters and freshwater bodies are severely challenged by high anthropogenic activities, complex hydrodynamics, and spatiotemporal variations of the environmental process. While advanced commercial robots and sensors are available for such a scenario, cost is often a concern. This presentation explores the use of simple sensors and robots with informative path planning as a more economically feasible alternative for large scale deployment. We present the robotic architecture, operation paradigm, and some robotic algorithms that we have experimented in a bid to improve the adoption of robotics in field operations. They include path planning algorithms to address the challenges in localisation, endurance, and environmental monitoring using small AUVs and miniature USVs. We discuss the online informative frameworks developed to perform in-situ monitoring and sample collection with bounds on the mission time. The performances of the paths generated by our frameworks are compared against conventional lawnmower paths using data from simulations and field experiments. The analysis of the samples collected showed the biological relevance of the field estimated using our adaptive sampling algorithms and that informed sampling can yield substantial information about the environment. The presentation then concludes with a preliminary work on addressing sensor needs and operational challenges in information exchange among heterogeneous assets.
Title: Multi-Objective Tree Search for Spatiotemporal Informative Planning
Abstract: Adaptive sampling and planning in robotic environmental monitoring are challenging when the target environmental process varies over space and time. The underlying environmental dynamics require the planning module to integrate future environmental changes so that action decisions made earlier do not quickly become outdated. I will discuss a Monte Carlo tree search method which not only well balances the environment exploration and exploitation in space, but also catches up to the environmental dynamics that are related to time. This is achieved by incorporating multi-objective optimization and a look-ahead model-predictive rewarding mechanism. The method produces optimized decision solutions for the robot based on its knowledge (estimation) of the environment state, leading to better adaptation to environmental dynamics.
Title: Creating Effective and Useful Robots for Scientific Data Collection
Abstract: This talk will discuss the challenges and issues that face researchers doing large scale data collection outdoors. I use examples from our lab to illustrate different kinds of approach that can be used, and different sensing contexts, as well as some of the open issues, difficulties and requirements such projects face. We have used a combination of autonomy, model-based reinforcement learning and human-driven in imitation learning. In work from Travis Manderson, we use such methods on both underwater and terrestrial robots. In work with Sandeep Manjanna, we focus on where to collect data for maximal efficiency. In work with Johanna Hansen, we exploit natural processes combined with learning to model complex natural phenomena. Part of my objective In this talk is also to expose issues for discussion in the subsequent panel.Back ↩