OpenSpiel: A Framework for Reinforcement Learning in Games

OpenSpiel is a collection of environments and algorithms for research in general reinforcement learning and search/planning in games. OpenSpiel supports n-player (single- and multi- agent) zero-sum, cooperative and general-sum, one-shot and sequential, strictly turn-taking and simultaneous-move, perfect and imperfect information games, as well as traditional multiagent environments such as (partially- and fully observable) grid worlds and social dilemmas. OpenSpiel also includes tools to analyze learning dynamics and other common evaluation metrics. This document serves both as an overview of the code base and an introduction to the terminology, core concepts, and algorithms across the fields of reinforcement learning, computational game theory, and search.

Neural Replicator Dynamics

In multiagent learning, agents interact in inherently nonstationary environments due to their concurrent policy updates. It is, therefore, paramount to develop and analyze algorithms that learn effectively despite these nonstationarities. A number of works have successfully conducted this analysis under the lens of evolutionary game theory (EGT), wherein a population of individuals interact and evolve based on biologically-inspired operators. These studies have mainly focused on establishing connections to value-iteration based approaches in stateless or tabular games. We extend this line of inquiry to formally establish links between EGT and policy gradient (PG) methods, which have been extensively applied in single and multiagent learning. We pinpoint weaknesses of the commonly-used softmax PG algorithm in adversarial and nonstationary settings and contrast PG's behavior to that predicted by replicator dynamics (RD), a central model in EGT. We consequently provide theoretical results that establish links between EGT and PG methods, then derive Neural Replicator Dynamics (NeuRD), a parameterized version of RD that constitutes a novel method with several advantages. First, as NeuRD reduces to the well-studied no-regret Hedge algorithm in the tabular setting, it inherits no-regret guarantees that enable convergence to equilibria in games. Second, NeuRD is shown to be more adaptive to nonstationarity, in comparison to PG, when learning in canonical games and imperfect information benchmarks including Poker. Thirdly, modifying any PG-based algorithm to use the NeuRD update rule is straightforward and incurs no added computational costs. Finally, while single-agent learning is not the main focus of the paper, we verify empirically that NeuRD is competitive in these settings with a recent baseline algorithm.

Differentiable Game Mechanics

Deep learning is built on the foundational guarantee that gradient descent on an objective function converges to local minima. Unfortunately, this guarantee fails in settings, such as generative adversarial nets, that exhibit multiple interacting losses. The behavior of gradient-based methods in games is not well understood -- and is becoming increasingly important as adversarial and multi-objective architectures proliferate. In this paper, we develop new tools to understand and control the dynamics in n-player differentiable games. The key result is to decompose the game Jacobian into two components. The first, symmetric component, is related to potential games, which reduce to gradient descent on an implicit function. The second, antisymmetric component, relates to Hamiltonian games, a new class of games that obey a conservation law akin to conservation laws in classical mechanical systems. The decomposition motivates Symplectic Gradient Adjustment (SGA), a new algorithm for finding stable fixed points in differentiable games. Basic experiments show SGA is competitive with recently proposed algorithms for finding stable fixed points in GANs -- while at the same time being applicable to, and having guarantees in, much more general cases.

Evolving Indoor Navigational Strategies Using Gated Recurrent Units In NEAT

Simultaneous Localisation and Mapping (SLAM) algorithms are expensive to run on smaller robotic platforms such as Micro-Aerial Vehicles. Bug algorithms are an alternative that use relatively little processing power, and avoid high memory consumption by not building an explicit map of the environment. Bug Algorithms achieve relatively good performance in simulated and robotic maze solving domains. However, because they are hand-designed, a natural question is whether they are globally optimal control policies. In this work we explore the performance of Neuroevolution - specifically NEAT - at evolving control policies for simulated differential drive robots carrying out generalised maze navigation. We extend NEAT to include Gated Recurrent Units (GRUs) to help deal with long term dependencies. We show that both NEAT and our NEAT-GRU can repeatably generate controllers that outperform I-Bug (an algorithm particularly well-suited for use in real robots) on a test set of 209 indoor maze like environments. We show that NEAT-GRU is superior to NEAT in this task but also that out of the 2 systems, only NEAT-GRU can continuously evolve successful controllers for a much harder task in which no bearing information about the target is provided to the agent.

Computing Approximate Equilibria in Sequential Adversarial Games by Exploitability Descent

In this paper, we present exploitability descent, a new algorithm to compute approximate equilibria in two-player zero-sum extensive-form games with imperfect information, by direct policy optimization against worst-case opponents. We prove that when following this optimization, the exploitability of a player's strategy converges asymptotically to zero, and hence when both players employ this optimization, the joint policies converge to a Nash equilibrium. Unlike fictitious play (XFP) and counterfactual regret minimization (CFR), our convergence result pertains to the policies being optimized rather than the average policies. Our experiments demonstrate convergence rates comparable to XFP and CFR in four benchmark games in the tabular case. Using function approximation, we find that our algorithm outperforms the tabular version in two of the games, which, to the best of our knowledge, is the first such result in imperfect information games among this class of algorithms.

Fully Convolutional One-Shot Object Segmentation for Industrial Robotics

The ability to identify and localize new objects robustly and effectively is vital for robotic grasping and manipulation in warehouses or smart factories. Deep convolutional neural networks (DCNNs) have achieved the state-of-the-art performance on established image datasets for object detection and segmentation. However, applying DCNNs in dynamic industrial scenarios, e.g., warehouses and autonomous production, remains a challenging problem. DCNNs quickly become ineffective when tasked with detecting objects that they have not been trained on. Given that re-training using the latest data is time consuming, DCNNs cannot meet the requirement of the Factory of the Future (FoF) regarding rapid development and production cycles. To address this problem, we propose a novel one-shot object segmentation framework, using a fully convolutional Siamese network architecture, to detect previously unknown objects based on a single prototype image. We turn to multi-task learning to reduce training time and improve classification accuracy. Furthermore, we introduce a novel approach to automatically cluster the learnt feature space representation in a weakly supervised manner. We test the proposed framework on the RoboCup@Work dataset, simulating requirements for the FoF. Results show that the trained network on average identifies 73% of previously unseen objects correctly from a single example image. Correctly identified objects are estimated to have a 87.53% successful pick-up rate. Finally, multi-task learning lowers the convergence time by up to 33%, and increases accuracy by 2.99%.

Robust temporal difference learning for critical domains

We present a new Q-function operator for temporal difference (TD) learning methods that explicitly encodes robustness against significant rare events (SRE) in critical domains. The operator, which we call the $\kappa$-operator, allows to learn a safe policy in a model-based fashion without actually observing the SRE. We introduce single- and multi-agent robust TD methods using the operator $\kappa$. We prove convergence of the operator to the optimal safe Q-function with respect to the model using the theory of Generalized Markov Decision Processes. In addition we prove convergence to the optimal Q-function of the original MDP given that the probability of SREs vanishes. Empirical evaluations demonstrate the superior performance of $\kappa$-based TD methods both in the early learning phase as well as in the final converged stage. In addition we show robustness of the proposed method to small model errors, as well as its applicability in a multi-agent context.

Re-evaluating Evaluation

Progress in machine learning is measured by careful evaluation on problems of outstanding common interest. However, the proliferation of benchmark suites and environments, adversarial attacks, and other complications has diluted the basic evaluation model by overwhelming researchers with choices. Deliberate or accidental cherry picking is increasingly likely, and designing well-balanced evaluation suites requires increasing effort. In this paper we take a step back and propose Nash averaging. The approach builds on a detailed analysis of the algebraic structure of evaluation in two basic scenarios: agent-vs-agent and agent-vs-task. The key strength of Nash averaging is that it automatically adapts to redundancies in evaluation data, so that results are not biased by the incorporation of easy tasks or weak agents. Nash averaging thus encourages maximally inclusive evaluation -- since there is no harm (computational cost aside) from including all available tasks and agents.

Actor-Critic Policy Optimization in Partially Observable Multiagent Environments

Optimization of parameterized policies for reinforcement learning (RL) is an important and challenging problem in artificial intelligence. Among the most common approaches are algorithms based on gradient ascent of a score function representing discounted return. In this paper, we examine the role of these policy gradient and actor-critic algorithms in partially-observable multiagent environments. We show several candidate policy update rules and relate them to a foundation of regret minimization and multiagent learning techniques for the one-shot and tabular cases, leading to previously unknown convergence guarantees. We apply our method to model-free multiagent reinforcement learning in adversarial sequential decision problems (zero-sum imperfect information games), using RL-style function approximation. We evaluate on commonly used benchmark Poker domains, showing performance against fixed policies and empirical convergence to approximate Nash equilibria in self-play with rates similar to or better than a baseline model-free algorithm for zero sum games, without any domain-specific state space reductions.

Inequity aversion improves cooperation in intertemporal social dilemmas

Groups of humans are often able to find ways to cooperate with one another in complex, temporally extended social dilemmas. Models based on behavioral economics are only able to explain this phenomenon for unrealistic stateless matrix games. Recently, multi-agent reinforcement learning has been applied to generalize social dilemma problems to temporally and spatially extended Markov games. However, this has not yet generated an agent that learns to cooperate in social dilemmas as humans do. A key insight is that many, but not all, human individuals have inequity averse social preferences. This promotes a particular resolution of the matrix game social dilemma wherein inequity-averse individuals are personally pro-social and punish defectors. Here we extend this idea to Markov games and show that it promotes cooperation in several types of sequential social dilemma, via a profitable interaction with policy learnability. In particular, we find that inequity aversion improves temporal credit assignment for the important class of intertemporal social dilemmas. These results help explain how large-scale cooperation may emerge and persist.