Automated Bidding

Automated bidding studies how advertisers and other agents choose bids across repeated auctions to win valuable opportunities while meeting budgets, return, or service constraints. Methods range from dual-based pacing and online bandit algorithms to model-predictive control, reinforcement learning, and generative policies, often using auction forecasts and safe fallback controls. Work spans first- and second-price advertising auctions as well as electricity, spectrum, procurement, and computing markets, evaluated with simulations, benchmarks, and online tests. A central challenge is adapting to changing or unreliable feedback while satisfying constraints; better predictions and more sophisticated agents do not automatically improve welfare, fairness, or platform outcomes.