Studying adversarial attacks on Reinforcement Learning (RL) agents has become a key aspect of developing robust, RL-based solutions. Test-time attacks, which target the post-learning performance of an RL agent's policy, have been well studied in both white- and black-box settings. More recently, however, state-of-the-art works have shifted to investigate training-time attacks on RL agents, i.e., forcing the learning process towards a target policy designed by the attacker. Alas, these SOTA works continue to rely on white-box settings and/or use a reward-poisoning approach. In contrast, this paper studies environment-dynamics poisoning attacks at training time. Furthermore, while environment-dynamics poisoning presumes a transfer-learning capable agent, it also allows us to expand our approach to black-box attacks. Our overall framework, inspired by hierarchical RL, seeks the minimal environment-dynamics manipulation that will prompt the momentary policy of the agent to change in a desired manner. We show the attack efficiency by comparing it with the reward-poisoning approach, and empirically demonstrate the transferability of the environment-poisoning attack strategy. Finally, we seek to exploit the transferability of the attack strategy to handle black-box settings.

Transferable Environment Poisoning: Training-time Attack on Reinforcement Learning

Distributed Constraint Optimization Problems (DCOPs) are a powerful tool to model multi-agent coordination problems that are distributed by nature. The formulation is suitable for problems where variables are discrete and constraint utilities are represented in tabular form. However, many real-world applications have variables that are continuous and tabular forms thus cannot accurately represent constraint utilities. To overcome this limitation, researchers have proposed the Continuous DCOP (C-DCOP) model, which are DCOPs with continuous variables. But existing approaches usually come with some restrictions on the form of constraint utilities and are without quality guarantees. Therefore, in this paper, we (i) propose an exact algorithm to solve a specific subclass of C-DCOPs; (ii) propose an approximation method with quality guarantees to solve general C-DCOPs; (iii) propose additional C-DCOP algorithms that are more scalable; and (v) empirically show that our algorithms outperform existing state-of-the-art C-DCOP algorithms when given the same communication limitations.

New Algorithms for Continuous Distributed Constraint Optimization Problems

This presentation introduces a teaching model BC-MDP to cultivate a population by tweaking the environment dynamics. The proposed BC-MDP combines the Behaviour Cultivation core with the recent advances of Curriculum MDPs. This allows BC-MDP to address several shortcomings of the classical BC, while preserving its strengths, such as the freedom from the teacher-learner value alignment. Our model exploits the knowledge of the learner population adaptation process to induce and proliferate a desired behaviour throughout the population. We experimentally show its effectiveness, and retention of key positive features of both BC and Curriculum-MDP.


Teaching Multiple Learning Agents by Environment-Dynamics Tweaks

S5: Multiagent Learning 3

technical paper

AAMAS is the largest and most influential conference in the area of agents and multiagent systems, bringing together researchers and practitioners in all areas of agent technology and providing and internationally renowned high-profile forum for publishing and finding out about the latest developments in the field. AAMAS is the flagship conference of the non-profit International Foundation for Autonomous Agents and Multiagent Systems (IFAAMAS).  

![](https://assets.underline.io/uploads/markdown_image/1/image/ce4604bd54468a006164d16e080204fb.jpg)

Please click on the button below to register for this event.

If you are registered and see this page please check the account email you are logged in with. It must be the same as for the registration.

If you still have trouble accessing the contact please contact support at aamas2021@underline.io 

AAMAS 2021

Knowledge Representation, Reasoning and Planning

AAMAS is the leading scientific conference for research in autonomous agents and multi-agent systems. The AAMAS conference series was initiated in 2002 as the merging of three respected scientific meetings: the International Conference on Multi-Agent Systems (ICMAS), the International Workshop on Agent Theories, Architectures, and Languages (ATAL), and the International Conference on Autonomous Agents (AA). The aim of the joint conference is to provide a single, high-profile, internationally-respected archival forum for scientific research in the theory and practice of autonomous agents and multi-agent systems.

Browse keynotes, discussions, panels and over 300 presentations.


AAMAS 2020

AAMAS is the leading scientific conference for research in autonomous agents and multi-agent systems. The AAMAS conference series was initiated in 2002 as the merging of three respected scientific meetings: the International Conference on Multi-Agent Systems (ICMAS), the International Workshop on Agent Theories, Architectures, and Languages (ATAL), and the International Conference on Autonomous Agents (AA). 

Zinovi Rabinovich

3

1

Presentations

Transferable Environment Poisoning: Training-time Attack on Reinforcement Learning

New Algorithms for Continuous Distributed Constraint Optimization Problems

Teaching Multiple Learning Agents by Environment-Dynamics Tweaks

Stay up to date with the latest Underline news!

PRESENTATIONS

CONFERENCES

COMPANY

RESOURCES