Dynamic Robotic Task Allocation using Real Time Path Planning with Field D*
Advisor Information
Prithviraj Dasgupta
Location
Milo Bail Student Center Council Room
Presentation Type
Oral Presentation
Start Date
8-3-2013 2:00 PM
End Date
8-3-2013 2:15 PM
Abstract
Multi -robot systems comprise of robot teams which operate collectively to perform the tasks that are assigned to them. One of the major problems encountered by multi -robot systems that operate in initially unknown or partially -known environments is that the conditions in the environment can change dynamically as the robots are performing the tasks. This makes it difficult for the robots to ‘plan’ their tasks ahead of time. However, most state -of -the -art task allocation techniques for multi -robot systems do not incorporate a real-time feedback about the situation of the environment encountered by the robots into their task planning. In this research, we propose to address this deficit by developing a two-layered approach - in the bottom layer, robots update their current perception of the environment, while, in the top -layer, they use the current perception to update their plan of performing tasks. We have demonstrated our technique in a scenario where the robots’ tasks correspond to visiting a set of locations and doing an operation at each location. Such scenarios are encountered in various robotics applications such as detecting potential landmines in an automated humanitarian demining scenario or putting out fires in an automated search and rescue scenario, etc. We have implemented our proposed technique on a Corobot robot within our lab environment and shown that the robot can dynamically detect and avoid obstacles and other mobile robots, and re -plan its path to visit its target locations while incorporating changes to its originally planned paths based on the dynamic and static obstacles it discovers .
Dynamic Robotic Task Allocation using Real Time Path Planning with Field D*
Milo Bail Student Center Council Room
Multi -robot systems comprise of robot teams which operate collectively to perform the tasks that are assigned to them. One of the major problems encountered by multi -robot systems that operate in initially unknown or partially -known environments is that the conditions in the environment can change dynamically as the robots are performing the tasks. This makes it difficult for the robots to ‘plan’ their tasks ahead of time. However, most state -of -the -art task allocation techniques for multi -robot systems do not incorporate a real-time feedback about the situation of the environment encountered by the robots into their task planning. In this research, we propose to address this deficit by developing a two-layered approach - in the bottom layer, robots update their current perception of the environment, while, in the top -layer, they use the current perception to update their plan of performing tasks. We have demonstrated our technique in a scenario where the robots’ tasks correspond to visiting a set of locations and doing an operation at each location. Such scenarios are encountered in various robotics applications such as detecting potential landmines in an automated humanitarian demining scenario or putting out fires in an automated search and rescue scenario, etc. We have implemented our proposed technique on a Corobot robot within our lab environment and shown that the robot can dynamically detect and avoid obstacles and other mobile robots, and re -plan its path to visit its target locations while incorporating changes to its originally planned paths based on the dynamic and static obstacles it discovers .