Deployment, verification, and optimization of algorithms for simultaneous task planning and allocation in dynamic environments.

Sammanfattning: With the upgrading of hardware such as sensors, processors, and artificial intelligence technologies, robotics technology is developing in the direction of intelligence. As multi-robot collaborative work is required in more and more scenarios, the robustness of multi-robot systems(MRS) against sensor errors, mapping errors, and dynamic obstacles is becoming a focus issue. An MRS framework named Simultaneous Planning and Allocation in Dynamic EnvironmentS(SPADES) is being developed to find a solution for operating multiple robots in dynamic indoor environments such as hospitals. It is a generally centralized system that consists of a central Task Allocation (TA) system combined with motion planners that run on individual robots. To further push it towards deployment on actual robots, three dimensions (3D) simulations of the SPADES framework were conducted, and the performance of the framework in the 3D simulations was compared with that of it in 2D simulations. Additional costs were introduced into the central task allocation system to deal with the uncertainty in the environment. In this paper, the background of robotics technology, the problems that the academic community faces, and the progress made in MRS are introduced. Then the constitute of the SPADES framework is explained from the algorithm level. After that, the implementation of the 3D simulation and the introduction of the uncertainty cost is described. The results of the comparison between the 3D and 2D simulations, 3D simulation under different influential factors, and the comparison of the framework before/after the optimization of uncertainty cost are demonstrated and analyzed. In conclusion, the spades framework performs poorer in 3D simulations, partly due to sensor noise and simulated physics effects. The introduction of the uncertainty costs can counter the negative influences brought by the dynamic environment to a certain degree