| 9:00 - 9:10 am |
Organizers |
Welcome and Introduction |
| 9:10 - 9:35 am |
Laura Fiorini
University of Florence
|
Personalized HRI through advanced behavioural models for application in healthcare domains: Where we are and what we need to |
Abstract: Social robots are entering our houses and hospitals, therefore they should be endowed with advanced and personalised interaction capabilities thus to adapt to the dynamic external environments and the changing needs of frail citizens. Additionally, social robots should be able to cooperate interactively with formal caregivers as well as clinicians thus shifting from a care model where the frail user interacts with the robot to care models where also the clinicians/formal caregivers are also involved at different levels, according to the application. Indeed, the data acquired from the robots can be used by the robots to personalise the interaction and by the clinician to monitor the status of the users. |
| 9:35 - 10:00 am
|
Teresa Vidal Calleja
University of Technology Sydney
| Physics driven, continuous and probabilistic representations for localisation, mapping and planning |
Abstract:
In this talk, first I will go through our work on faithful Euclidean distance field estimation for localisation, mapping and planning using a continuous and probabilistic implicit surface representation (Log-GPIS). Log-GPIS aims to approximate closely the solution of the regularised Eikonal equation to estimate the distance field and its gradient enabling surface reconstruction, localisation and obstacle avoidance. Then, I will introduce our recent work on global localisation based on continuous magnetic vector fields that rely on a divergence-free kernel and our crowd prediction approach that enforces the conservation of people density. Simulations and experimental results will be used to show the performance of these representations.
|
| 10:00 - 10:25 am
|
Hyun-Taek Choi
Korea Research Institute of Ships and Ocean Engineering
|
Understanding and Representing the Sea environment for Autonomous Ship Navigation |
Abstract: Navigation in robotics has seen significant progress in recent years, with advancements in various technologies such as SLAM (Simultaneous Localization and Mapping). However, implementing these technologies in real environments for extended periods of time with both robustness and flexibility remains a challenge. It is possible that these established frameworks may be leading us to overlook real problems. Although the ocean may not be appropriate for pure SLAM, by understanding the environment and our goals precisely, the problem we need to solve does not have to be unnecessarily difficult. In this context, the environment includes not only wide areas of the ocean with strong sunlight and dense fog but also the temporal conditions of control cycles necessary for ship navigation, peripheral conditions that change according to the navigation route of the ship, and the conditions imposed by international law and conventions in the shipping industry. This presentation introduces our ongoing research on the situation awareness system of autonomous ships, which is being developed due to commercial demand. We will show you our data fusion structure using mathematical and heuristic methods, along with deep-learning-based detection algorithms for each sensor. In conclusion, we keep reminding ourselves that our goal is to identify any collision risks rather than detecting small objects near the horizon. To successfully apply autonomous systems in various applications, it is beneficial to take a heuristic approach, accurately understanding the system and the purpose we pursue.
|
| 10:25 - 10:50 am |
Poster session and coffee break |
| 10:50 - 11:35 am |
"World Cafe" discussion |
- What are you missing from available maps representations?
- What would be a significant breakthrough to enable a wide application of autonomous mobile robots in broader applications fields
- How can the community benefit from some "unification" processes to allow the mixing of various spatial representations with semantic interpretations?
- Is there a need for an international standard representation?
|
| 11:35 - 12:00 pm
|
Yvan Petillot
Heriot-Watt University
| Map representations for remote marine operations, what can we get? What do we need? How do we get there? |
Abstract: TBA |
| 12:00 - 12:25 pm
|
Anna Mannucci & Luigi Palmieri
Robert Bosch GmbH
|
Towards Context-aware Predictive Planning in Complex Environments |
Abstract: Computing safe navigation policies for wheeled mobile robots navigating in densely cluttered and crowded spaces is a difficult task due to several factors, e.g., perception noise, system models mismatch, high uncertainty of human future behaviors. The latter being influenced not only by other surrounding humans but also by environmental properties. In these settings, classical reactive approaches often result in an overly cautious robot that fails to produce a feasible, safe path in the crowd, or plans a large, sub-optimal, perhaps oscillating detour to avoid hindrances. Additionally, several contextual cues may influence robots' motion, e.g., semantic relationships between objects in the environment, activity patterns and social relations: considering them in the decision-making phase is fundamental for improving the overall robot operation efficiency. Due to those several factors, a unique solution to fully solve robot navigation in cluttered, crowded and dynamic environments is still far ahead of us. The problem is even more challenging when considering fleets of robots. In this talk, we will present several predictive robot motion planning approaches and architectures developed to solve the issue. Particularly demanding is the type of environment representations used in those architectures for computing the final robot policies.
We will show how the quest for a safe and efficient robot navigation policy requires not only the improvements of several planning sub-components, but also the study of proper architectures that consider contextual proprieties of the environment. |
| 12:25 - 1:55 apm |
Lunch Break |
| 1:55 - 2:20 pm
|
Jaeho Lee
|
Cloud Robotics |
Abstract: TBA |
| 2:20 - 2:45 pm
|
Edson Prestes
Universidade Federal do Rio Grande do Sul
|
IEEE Ontological Standard for Ethically Driven Robotics and Automation Systems |
Abstract: Artificial Intelligence and Robotics can bring many benefits to humanity. There are several examples created by our community that show how AI and robotics can be used to attain the United Nations Sustainable Development Goals. However, despite all these benefits, there are some applications that show the devastating power of AI and Robotics,
posing serious risks to our fundamental rights that go beyond privacy issues and that mainly affect vulnerable and underprivileged communities. Therefore, global society has put a lot of energy in creating soft and hard laws to create some barriers in these developments to ensure that AI and robotics based applications are used for good and not the other way around.
In this talk, I will discuss the recently published "IEEE Ontological Standard for Ethically Driven Robotics and Automation Systems" which aims mainly to assist in the development of ethically oriented methodologies for the design of robots and automation systems. However, this standard can be used in many ways, for example as the core of a platform for multilateral organisations to govern the domain.
|
| 2:45 - 3:10 pm
|
Hiroshi Ishida
Tokyo University of Agriculture
Haruka Matsukura
University of Electro-Communications
|
Olfactory landscapes of the world and challenges to digitize them
|
Abstract:
Olfaction is the sense of smell that enables organisms to detect volatile chemical compounds in the air and analyze their implications, such as food rottenness and homing orientation. Although attempts to provide robots with such a sensor modality have not always been successful, an increasing number of research efforts are being made as a result of advances in sensor technologies. Smell source localization has been one of the main topics in mobile robot olfaction. Some animals can locate food by tracking its smell. To accomplish smell tracking is very challenging because the aerial trail of smell is highly unstable. However, some successful results of gas source localization are reported using stochastic approaches, e.g., a particle filter. This talk also covers some recent topics including super-resolution for gas distribution mapping and digital reproduction of smells. |
| 3:10 - 3:35 pm
|
Ryan Smith
Fort Lewis College
|
Hard Miles without Hard Miles |
Abstract: It is miserable and costly to drive vast numbers of miles in the hope of experiencing those elusive edge/corner cases. Indeed, in many industrial and urban application domains, it is not even possible to do this in advance of substantive deployments. We will offer an alternative, less-dreary vista. Using composite scene synthesis and weather synthesis, software-in-the-loop tightly integrated with Sim, and reinforcement learning, we exponentiate the value of a small-seed dataset of benign autonomy runs as a precursor for a site-wide/domain-specific validation and verification. |
| 3:35 - 4:05 pm |
Poster session and coffee break |
| 4:05 - 4:50 pm |
"World Cafe" discussion |
- What are you missing from available maps representations?
- What would be a significant breakthrough to enable a wide application of autonomous mobile robots in broader applications fields
- How can the community benefit from some "unification" processes to allow the mixing of various spatial representations with semantic interpretations?
- Is there a need for an international standard representation?
|
| 4:50 - 5:00 pm |
Organizers |
Closing |
