Auction-Based Online Policy Adaptation for Evolving Objectives

TL;DR

This paper introduces an auction-based modular framework for multi-objective reinforcement learning that adapts dynamically to changing objectives, enabling efficient and interpretable policy adjustments in real-time.

Contribution

It proposes a novel auction mechanism for coordinating local policies in multi-objective RL, allowing immediate adaptation and improved performance over monolithic policies.

Findings

The method outperforms monolithic PPO-trained policies in experiments.

Policies can adapt instantly to objective changes by adding or removing local policies.

Theoretical analysis guarantees the existence of Nash equilibria with desirable properties.

Abstract

We consider multi-objective reinforcement learning problems where objectives come from an identical family -- such as the class of reachability objectives -- and may appear or disappear at runtime. Our goal is to design adaptive policies that can efficiently adjust their behaviors as the set of active objectives changes. To solve this problem, we propose a modular framework where each objective is supported by a selfish local policy, and coordination is achieved through a novel auction-based mechanism: policies bid for the right to execute their actions, with bids reflecting the urgency of the current state. The highest bidder selects the action, enabling a dynamic and interpretable trade-off among objectives. Going back to the original adaptation problem, when objectives change, the system adapts by simply adding or removing the corresponding policies. Moreover, as objectives arise from the same family, identical copies of a parameterized policy can be deployed, facilitating immediate adaptation at runtime. We show how the selfish local policies can be computed by turning the problem into a general-sum Markov game, where the policies compete against each other to fulfill their own objectives. To succeed, each policy must not only optimize its own objective, but also reason about the presence of other goals and learn to produce calibrated bids that reflect relative priority. Under mild assumptions, we prove the existence of Nash equilibria where dishonest bidding leads to suboptimal outcome, and the most urgent objectives win control automatically. In our implementation, the policies are trained concurrently using proximal policy optimization (PPO). We evaluate on two Atari games and a gridworld-based path-planning task with dynamic targets. Our method achieves substantially better performance than monolithic policies trained with PPO.