In order for agents trained by deep reinforcement learning to work alongside humans in realistic settings, we will need to ensure that the agents are \emph{robust}. Since the real world is very diverse, and human behavior often changes in response to agent deployment, the agent will likely encounter novel situations that have never been seen during training. This results in an evaluation challenge: if we cannot rely on the average training or validation reward as a metric, then how can we effectively evaluate robustness? We take inspiration from the practice of \emph{unit testing} in software engineering. Specifically, we suggest that when designing AI agents that collaborate with humans, designers should search for potential edge cases in \emph{possible partner behavior} and \emph{possible states encountered}, and write tests which check that the behavior of the agent in these edge cases is reasonable. We apply this methodology to build a suite of unit tests for the Overcooked-AI environment, and use this test suite to evaluate three proposals for improving robustness. We find that the test suite provides significant insight into the effects of these proposals that were generally not revealed by looking solely at the average validation reward.

We present a scalable tree search planning algorithm for large multi-agent sequential decision problems that require dynamic collaboration. Teams of agents need to coordinate decisions in many domains, but naive approaches fail due to the exponential growth of the joint action space with the number of agents. We circumvent this complexity through an anytime approach that allows us to trade computation for approximation quality and also dynamically coordinate actions. Our algorithm comprises three elements: online planning with Monte Carlo Tree Search (MCTS), factored representations of local agent interactions with coordination graphs, and the iterative Max-Plus method for joint action selection. We evaluate our approach on the benchmark SysAdmin domain with static coordination graphs and achieve comparable performance with much lower computation cost than our MCTS baselines. We also introduce a multi-drone delivery domain with dynamic, i.e., state-dependent coordination graphs, and demonstrate how our approach scales to large problems on this domain that are intractable for other MCTS methods. We provide an open-source implementation of our algorithm at https://github.com/JuliaPOMDP/FactoredValueMCTS.jl.

This paper presents a computational evolutionary game model to study and understand fraud dynamics in the consumption tax system. Players are cooperators if they correctly declare their value added tax (VAT), and are defectors otherwise. Each player's payoff is influenced by the amount evaded and the subjective probability of being inspected by tax authorities. Since transactions between companies must be declared by both the buyer and seller, a strategy adopted by one influences the other's payoff. We study the model with a well-mixed population and different scale-free networks. Model parameters were calibrated using real-world data of VAT declarations by businesses registered in the Canary Islands region of Spain. We analyzed several scenarios of audit probabilities for high and low transactions and their prevalence in the population, as well as social rewards and penalties to find the most efficient policy to increase the proportion of cooperators. Two major insights were found. First, increasing the subjective audit probability for low transactions is more efficient than increasing this probability for high transactions. Second, favoring social rewards for cooperators or alternative penalties for defectors can be effective policies, but their success depends on the distribution of the audit probability for low and high transactions.

We study the problem of distributed multi-robot coverage over an unknown, nonuniform sensory field. Modeling the sensory field as a realization of a Gaussian Process and using Bayesian techniques, we devise a policy which aims to balance the tradeoff between learning the sensory function and covering the environment. We propose an adaptive coverage algorithm called Deterministic Sequencing of Learning and Coverage (DSLC) that schedules learning and coverage epochs such that its emphasis gradually shifts from exploration to exploitation while never fully ceasing to learn. Using a novel definition of coverage regret which characterizes overall coverage performance of a multi-robot team over a time horizon $T$, we analyze DSLC to provide an upper bound on expected cumulative coverage regret. Finally, we illustrate the empirical performance of the algorithm through simulations of the coverage task over an unknown distribution of wildfires.

Conventional multi-agent path planners typically compute an ensemble of paths while optimizing a single objective, such as path length. However, many applications may require multiple objectives, say fuel consumption and completion time, to be simultaneously optimized during planning and these criteria may not be readily compared and sometimes lie in competition with each other. Naively applying existing multi-objective search algorithms to multi-agent path finding may prove to be inefficient as the size of the space of possible solutions, i.e., the Pareto-optimal set, can grow exponentially with the number of agents (the dimension of the search space). This article presents an approach named Multi-objective Conflict-based Search (MO-CBS) that bypasses this so-called curse of dimensionality by leveraging prior Conflict-based Search (CBS), a well-known algorithm for single-objective multi-agent path finding, and principles of dominance from multi-objective optimization literature. We prove that MO-CBS is able to compute the entire Pareto-optimal set. Our results show that MO-CBS can solve problem instances with hundreds of Pareto-optimal solutions which the standard multi-objective A* algorithms could not find within a bounded time.

We address the packet routing problem in highly dynamic mobile ad-hoc networks (MANETs). In the network routing problem each router chooses the next-hop(s) of each packet to deliver the packet to a destination with lower delay, higher reliability, and less overhead in the network. In this paper, we present a novel framework and routing policies, DeepCQ+ routing, using multi-agent deep reinforcement learning (MADRL) which is designed to be robust and scalable for MANETs. Unlike other deep reinforcement learning (DRL)-based routing solutions in the literature, our approach has enabled us to train over a limited range of network parameters and conditions, but achieve realistic routing policies for a much wider range of conditions including a variable number of nodes, different data flows with varying data rates and source/destination pairs, diverse mobility levels, and other dynamic topology of networks. We demonstrate the scalability, robustness, and performance enhancements obtained by DeepCQ+ routing over a recently proposed model-free and non-neural robust and reliable routing technique (i.e. CQ+ routing). DeepCQ+ routing outperforms non-DRL-based CQ+ routing in terms of overhead while maintains same goodput rate. Under a wide range of network sizes and mobility conditions, we have observed the reduction in normalized overhead of 10-15%, indicating that the DeepCQ+ routing policy delivers more packets end-to-end with less overhead used. To the best of our knowledge, this is the first successful application of MADRL for the MANET routing problem that simultaneously achieves scalability and robustness under dynamic conditions while outperforming its non-neural counterpart. More importantly, we provide a framework to design scalable and robust routing policy with any desired network performance metric of interest.

IJCAI-PRICAI2020, the 29th International Joint Conference on Artificial Intelligence and the 17th Pacific Rim International Conference on Artificial Intelligence starts today and will run until 15 January. Find out what's happening during the event. The conference schedule is here and includes tutorials, workshops, invited talks and technical sessions. There are also competitions, early career spotlight talks, panel discussions and social events. There will be eight invited talks on a wide variety of topics.

Artificial Intelligence (AI) is being increasingly deployed in practical applications. However, there is a major concern whether AI systems will be trusted by humans. In order to establish trust in AI systems, there is a need for users to understand the reasoning behind their solutions. Therefore, systems should be able to explain and justify their output. In this paper, we propose an argument scheme-based approach to provide explanations in the domain of AI planning. We present novel argument schemes to create arguments that explain a plan and its key elements; and a set of critical questions that allow interaction between the arguments and enable the user to obtain further information regarding the key elements of the plan. Furthermore, we present a novel dialogue system using the argument schemes and critical questions for providing interactive dialectical explanations.

Sender-receiver interactions, and specifically persuasion games, are widely researched in economic modeling and artificial intelligence, and serve as a solid foundation for powerful applications. However, in the classic persuasion games setting, the messages sent from the expert to the decision-maker are abstract or well-structured application-specific signals rather than natural (human) language messages, although natural language is a very common communication signal in real-world persuasion setups. This paper addresses the use of natural language in persuasion games, exploring its impact on the decisions made by the players and aiming to construct effective models for the prediction of these decisions. For this purpose, we conduct an online repeated interaction experiment. At each trial of the interaction, an informed expert aims to sell an uninformed decision-maker a vacation in a hotel, by sending her a review that describes the hotel. While the expert is exposed to several scored reviews, the decision-maker observes only the single review sent by the expert, and her payoff in case she chooses to take the hotel is a random draw from the review score distribution available to the expert only. The expert's payoff, in turn, depends on the number of times the decision-maker chooses the hotel. We consider a number of modeling approaches for this setup, differing from each other in the model type (deep neural network (DNN) vs. linear classifier), the type of features used by the model (textual, behavioral or both) and the source of the textual features (DNN-based vs. hand-crafted). Our results demonstrate that given a prefix of the interaction sequence, our models can predict the future decisions of the decision-maker, particularly when a sequential modeling approach and hand-crafted textual features are applied.

Simulations, along with other similar applications like virtual worlds and video games, require computational models of intelligence that generate realistic and credible behavior for the participating synthetic characters. Cognitive architectures, which are models of the fixed structure underlying intelligent behavior in both natural and artificial systems, provide a conceptually valid common basis, as evidenced by the current efforts towards a standard model of the mind, to generate human-like intelligent behavior for these synthetic characters. Sigma is a cognitive architecture and system that strives to combine what has been learned from four decades of independent work on symbolic cognitive architectures, probabilistic graphical models, and more recently neural models, under its graphical architecture hypothesis. Sigma leverages an extended form of factor graphs towards a uniform grand unification of not only traditional cognitive capabilities but also key non-cognitive aspects, creating unique opportunities for the construction of new kinds of cognitive models that possess a Theory-of-Mind and that are perceptual, autonomous, interactive, affective, and adaptive. In this paper, we will introduce Sigma along with its diverse capabilities and then use three distinct proof-of-concept Sigma models to highlight combinations of these capabilities: (1) Distributional reinforcement learning models in; (2) A pair of adaptive and interactive agent models that demonstrate rule-based, probabilistic, and social reasoning; and (3) A knowledge-free exploration model in which an agent leverages only architectural appraisal variables, namely attention and curiosity, to locate an item while building up a map in a Unity environment.