A Formal Framework for Reactive Heterogeneous Multirobot Task Allocation in Uncertain Semantic Environments.

Area of Science:

• Robotics
• Artificial Intelligence
• Planning and Scheduling

Background:

• Current multi-robot task allocation (MRTA) often assumes complete environmental knowledge.
• Real-world scenarios present semantic uncertainties and dynamic task requirements, hindering reactive robot responses.
• Tasks frequently involve complex constraints like temporal needs, diverse robot capabilities, resource demands, and inter-task dependencies.

Purpose of the Study:

• To develop a reactive task allocation framework for heterogeneous multirobot systems.
• To address temporal requirements and multiple task constraints using Linear Temporal Logic with Rewards ($\mathrm {LTL^{\mathcal {R}}}$).
• To manage environments with known geometry but unknown semantic landmarks.

Main Methods:

• Proposed a reactive multiconstraint planning decision tree (RMC-PDT) for efficient task allocation.
• Encoded $\mathrm {LTL^{\mathcal {R}}}$ specifications and system states into the planning decision tree.
• Implemented a re-planning mechanism triggered by the detection of relevant semantic landmarks.

Main Results:

• Demonstrated efficient reactive planning capabilities in dynamic environments.
• Successfully solved complex task allocations involving multiple constraints.
• Validated the scalability of the proposed framework through extensive experiments.

Conclusions:

• The RMC-PDT framework effectively addresses uncertainties and dynamic changes in multirobot task allocation.
• The approach provides efficient, scalable, and constraint-aware solutions for heterogeneous multirobot systems.
• This work advances reactive planning for robots operating in complex, partially unknown environments.