Extending Consensus-based Task Allocation Algorithms with Bid Intercession to Foster Mixed-Initiative

We propose a new approach for controlling task allocation in teams of robots with different capabilities. This approach allows human operators, who have a better understanding of the situation, to influence or even dictate how tasks are distributed, whilst allowing autonomous decisions. Our method works within existing consensusbased allocation algorithms by introducing intercession in the bidding process. Intercession allows agents to bid on behalf of others. This allows for a flexible range of control, from completely decentralized to fully human-controlled, without refactoring the consensus-based allocation scheme, which has been proven to be efficient. We build upon an existing algorithm, Consensus-based Bundle Auction (CBBA), while maintaining its solution quality and ability to reach agreement (convergence). We test our new method, I-CBBA, in simulated multi-robot task allocation (MRTA) scenarios using the ROS framework.