Auction-Based Task Allocation Under Untruthful Agents

In this paper, we propose an iterative task negotiation algorithm for distributed multi-agent systems in the presence of deceptive agents. Compared to the baseline consensus-based bundle algorithm, the proposed method integrates a second-price payment scheme that incentivizes truthful bidding among honest agents. This framework can capture both truthful behavior and strategic deception, wherein agents may overbid to secure additional utility regardless of payment. The auction-based allocation algorithm is embedded within an iterative negotiation loop. In each iteration, agents estimate the trustworthiness of their opponents using Bayesian inference, which in turn determines the extent of task information to disclose. This enables agents to limit negotiation with deceptive counterparts while maintaining full engagement with trustworthy ones by adaptively adjusting the accessible task set based on opponent type. The effectiveness of the proposed algorithm is demonstrated in a 2D multi-agent task allocation scenario involving varying degrees of deception.