Opportunistic Competition and Collaboration in Two-Robot Teams
Advisor Information
Prithviraj Dasgupta
Location
UNO Criss Library, Room 231
Presentation Type
Oral Presentation
Start Date
6-3-2015 10:00 AM
End Date
6-3-2015 10:15 AM
Abstract
In my research I have investigated the problem of autonomous coordination between robots, where each robot has to decide whether to work cooperatively or competitively to accomplish a set of assigned tasks. A task for a robot is abstracted as a location in the robot’s environment that must be visited to perform certain operations, such as collecting samples. The robot’s objective is to maximize the mass of samples collected while minimizing the amount of fuel spent. Each robot uses a graph theoretic algorithm, called Dijkstra’s algorithm, to find the shortest cost path for visiting the task locations. The cost is measured in terms of mass collected and fuel expended. Each robot then examines whether its cost is lower when visiting tasks individually or when collaborating and visiting tasks together. If both robots have lower individual path costs than the collaborative cost, they act individually. Conversely, if the collaborative cost is smaller than either individual cost, they collaborate. In the event of conflict, when one robot’s individual cost is smaller than the collaboration cost and the other robot’s cost is larger, the concept of maximizing social benefit in a group was used, meaning: the robots decide to act collaboratively only if doing so results in lower combined costs than the combined cost when acting individually. The proposed solution was implemented on the SPHERES robots within the Zero Robotics simulator, which is used for the MIT/NASA Zero Robotics Challenge. The results from simulating different two robot scenarios with different task locations and sample mass, will be demonstrated during the presentation.
Opportunistic Competition and Collaboration in Two-Robot Teams
UNO Criss Library, Room 231
In my research I have investigated the problem of autonomous coordination between robots, where each robot has to decide whether to work cooperatively or competitively to accomplish a set of assigned tasks. A task for a robot is abstracted as a location in the robot’s environment that must be visited to perform certain operations, such as collecting samples. The robot’s objective is to maximize the mass of samples collected while minimizing the amount of fuel spent. Each robot uses a graph theoretic algorithm, called Dijkstra’s algorithm, to find the shortest cost path for visiting the task locations. The cost is measured in terms of mass collected and fuel expended. Each robot then examines whether its cost is lower when visiting tasks individually or when collaborating and visiting tasks together. If both robots have lower individual path costs than the collaborative cost, they act individually. Conversely, if the collaborative cost is smaller than either individual cost, they collaborate. In the event of conflict, when one robot’s individual cost is smaller than the collaboration cost and the other robot’s cost is larger, the concept of maximizing social benefit in a group was used, meaning: the robots decide to act collaboratively only if doing so results in lower combined costs than the combined cost when acting individually. The proposed solution was implemented on the SPHERES robots within the Zero Robotics simulator, which is used for the MIT/NASA Zero Robotics Challenge. The results from simulating different two robot scenarios with different task locations and sample mass, will be demonstrated during the presentation.