Loading...
Collaborative perception for multi-robot team awareness
Gao, Peng
Gao, Peng
Citations
Altmetric:
Advisor
Editor
Date
Date Issued
2022
Date Submitted
Keywords
Collections
Research Projects
Organizational Units
Journal Issue
Embargo Expires
Abstract
Nowadays, multi-robot systems are becoming essential in many areas of modern society due to their reliability, parallelism, effectiveness, and flexibility to address collaborative tasks, such as human-autonomy collaborative manufacturing, multi-robot search and rescue, connected autonomous driving, and multi-drone delivery services. With the development of autonomous robots and the increasing complexity of applications, multi-robot systems become entrenched in our lives.
To enable intelligent multi-robot systems, team awareness is critical for a team of robots to improve shared understanding of their environments, as well as for individual robots to understand their teammates. The goal of team awareness is to allow robots to understand both unstructured environments and their teammates so that multiple robots can build shared situational awareness and collaboratively complete complex tasks as a team. However, it is extremely challenging to construct team awareness in multi-robot teaming due to the high dynamics and uncertainty of the environments, as well as complex interactions between robots and humans. To address challenges like these, collaborative perception in multi-robot systems is needed, which allows robots to represent, associate, integrate, and utilize individual robots' situational awareness to build the shared, robust, and efficient team awareness for complex task completion.
This dissertation research mainly focuses on enabling multi-robot collaborative perception, which has addressed several important aspects of team awareness. First, I developed several correspondence identification methods that identify correspondences of the same objects in observations acquired by multiple teammates, in order to allow robots to have consistent visual references of these objects. Second, I introduced several multi-robot sensor fusion algorithms that integrate observations obtained by a team of robots, for the purpose of improving individual robots' situational awareness. Third, I realized approaches of heterogeneous multi-robot collaborative scheduling based on multi-robot team awareness, which allows optimal task allocations and autonomous adaptation to robot failures. These proposed methods do not just address specific tasks such as removing blind spots, improving object localization and collaborative scheduling, and application scenarios such as human-robot collaborative assembly, connected vehicle systems, autonomous driving, and robot-assisted manufacturing, but conclusively show the value of collaborative perception that can broaden the scope of multi-robot collaboration and enable multi-robot collective intelligence.
Associated Publications
Rights
Copyright of the original work is retained by the author.