Final Report Summary - COOPNET (Cooperative Situational Awareness for Wireless Networks)
The ERC Starting Grant project COOPNET started from the observation that future wireless networks will face fundamental challenges in the evolution of communication, sensing, timekeeping, and positioning. In particular, new ways of designing the associated systems are needed to maintain progress in terms of quality of service for each of these areas, aligned with the needs of future applications. The COOPNET project evaluated two related hypotheses:
1. Each of these areas can benefit significantly from cooperation among wireless nodes;
2. When cooperation is enabled, this leads to new and unforeseen trade-offs.
In order to evaluate these hypotheses, we investigated the interplay between localization, timekeeping, communication, and sensing in cooperative wireless systems. In such systems, agents (e.g. phones, robots, cars) help each other to achieve a common objective. This helping often requires some form of situational awareness (positioning and timekeeping) for the agents to know where they are and what/who is around them right now. In addition, helping is achieved through explicit wireless communication. Finally, sensing is needed to explore the environment and collect data. These three tasks are closely related, though they are often considered independently in the research community. Our research aims to reveal connections between these tasks, their trade-offs, and methods to harness their connections. This has led to new methods for cooperative systems, for instance allowing phones to cooperate in predicting link qualities based on their positions, and for providing fast location information for very large and dense networks. Finally, we have developed a multi-robot testbed, which allows us to validate these methods. Applications of this work includes 5G networks, autonomous robots, and cooperative intelligent transportation systems.
Our studies spanned theoretical contributions in the form of (i) fundamental performance bounds and intrinsic trade-offs, (ii) practical algorithms that are amenable for distributed processing, which operate close to the bounds and are able to harness the trade-offs, and (iii) validation through the testbed. From our studies, we found that:
1.Localization timekeeping, communication, and sensing should be treated together, not separately, in order for a design to be reasonably corresponding to reality. We have developed suitable tools that can support such designs and derived fundamental bounds and trade-offs.
2.Location uncertainty has a significant impact in certain communication and sensing applications, and cannot be ignored. We have developed machine learning tools that are able to incorporate uncertainty in communication and sensing.
3.Cooperating is not always beneficial for a multi-agent system, when considering the conflicting demands of localization, communication, and sensing. In particular, care should be taken whom to cooperative with. Our analysis quantifies these benefits.
These studies have been published in 3 licentiate theses, 41 peer-reviewed conference contributions, and 14 top-tier journal papers.
Our testbed had as original intention to provide a web-based interface to support researchers around the world, by allowing remote access to our facilities, run experiments, gather data, and validate algorithms. This turned out to be far more challenging than originally planned, as such a testbed requires human supervision while running, and would thus require around the clock physical support of the testbed. In addition, long-term lab space, required for the planned set-up, was not available at the host institution. The testbed is currently operational and we provide physical access to anyone in the research community to visit us and run experiments or collect data. While this limits the access, we have hosted many visiting scholars who have used the testbed in some form.
Finally, we also mention how the COOPNET research will continue beyond the lifetime of the ERC project. Our research has gathered attention from the control theory community, which has traditionally utilized overly simplified models of localization and communication. Now that the time has come for various types of cooperating agents (UAVs, robots, cars), COOPNET research has found a natural application. In addition, there is a need for integration of the COOPNET research with control and planning.
1. Each of these areas can benefit significantly from cooperation among wireless nodes;
2. When cooperation is enabled, this leads to new and unforeseen trade-offs.
In order to evaluate these hypotheses, we investigated the interplay between localization, timekeeping, communication, and sensing in cooperative wireless systems. In such systems, agents (e.g. phones, robots, cars) help each other to achieve a common objective. This helping often requires some form of situational awareness (positioning and timekeeping) for the agents to know where they are and what/who is around them right now. In addition, helping is achieved through explicit wireless communication. Finally, sensing is needed to explore the environment and collect data. These three tasks are closely related, though they are often considered independently in the research community. Our research aims to reveal connections between these tasks, their trade-offs, and methods to harness their connections. This has led to new methods for cooperative systems, for instance allowing phones to cooperate in predicting link qualities based on their positions, and for providing fast location information for very large and dense networks. Finally, we have developed a multi-robot testbed, which allows us to validate these methods. Applications of this work includes 5G networks, autonomous robots, and cooperative intelligent transportation systems.
Our studies spanned theoretical contributions in the form of (i) fundamental performance bounds and intrinsic trade-offs, (ii) practical algorithms that are amenable for distributed processing, which operate close to the bounds and are able to harness the trade-offs, and (iii) validation through the testbed. From our studies, we found that:
1.Localization timekeeping, communication, and sensing should be treated together, not separately, in order for a design to be reasonably corresponding to reality. We have developed suitable tools that can support such designs and derived fundamental bounds and trade-offs.
2.Location uncertainty has a significant impact in certain communication and sensing applications, and cannot be ignored. We have developed machine learning tools that are able to incorporate uncertainty in communication and sensing.
3.Cooperating is not always beneficial for a multi-agent system, when considering the conflicting demands of localization, communication, and sensing. In particular, care should be taken whom to cooperative with. Our analysis quantifies these benefits.
These studies have been published in 3 licentiate theses, 41 peer-reviewed conference contributions, and 14 top-tier journal papers.
Our testbed had as original intention to provide a web-based interface to support researchers around the world, by allowing remote access to our facilities, run experiments, gather data, and validate algorithms. This turned out to be far more challenging than originally planned, as such a testbed requires human supervision while running, and would thus require around the clock physical support of the testbed. In addition, long-term lab space, required for the planned set-up, was not available at the host institution. The testbed is currently operational and we provide physical access to anyone in the research community to visit us and run experiments or collect data. While this limits the access, we have hosted many visiting scholars who have used the testbed in some form.
Finally, we also mention how the COOPNET research will continue beyond the lifetime of the ERC project. Our research has gathered attention from the control theory community, which has traditionally utilized overly simplified models of localization and communication. Now that the time has come for various types of cooperating agents (UAVs, robots, cars), COOPNET research has found a natural application. In addition, there is a need for integration of the COOPNET research with control and planning.