-- We present Interactive Gibson, the first comprehensive benchmark for training and evaluating Interactive Navigation: robot navigation strategies where physical interaction with objects is allowed and even encouraged to accomplish a task. For example, the robot can move objects if needed in order to clear a path leading to the goal location. Our benchmark comprises two novel elements: 1) a new experimental setup, the Interactive Gibson Environment, which simulates high fidelity visuals of indoor scenes, and high fidelity physical dynamics of the robot and common objects found in these scenes; 2) a set of Interactive Navigation metrics which allows one to study the interplay between navigation and physical interaction. We present and evaluate multiple learning-based baselines in Interactive Gibson, and provide insights into regimes of navigation with different tradeoffs between navigation path efficiency and disturbance of surrounding objects. Classical robot navigation is concerned with reaching goals while avoiding collisions [1], [2]. This definition of navigation is motivated by a wide variety of robot applications in factories or outdoor settings. As robots are increasingly deployed in complex and cluttered environments, physical interactions while navigating become not only unavoidable, but necessary. For example, when operating a robot in a cluttered home, the robot might need to push objects aside or open doors in order to be able to reach its destination. This problem is referred to as Interactive Navigation and in this paper we propose a principled and systematic way to study it (see Figure 1). The "aversion to interaction" in robot mobile agents is easy to understand: real robots are expensive, and interacting with the environment presents safety risks. In Robotic Manipulation these challenges have been addressed by extensive use of physics simulation engines [3], [4], [5], which simulate object and robot dynamics with high precision and thus allow one to study manipulation in a safe manner. Further, these engines can be used to train models which are deployable in the real world.

Automatic T esting and Falsification with Dynamically Constrained Reinforcement Learning Xin Qin 1, Nikos Ar echiga 2, Andrew Best 2, Jyotirmoy Deshmukh 1 Abstract -- We consider the problem of using reinforcement learning to train adversarial agents for automatic testing and falsification of cyberphysical systems, such as autonomous vehicles, robots, and airplanes. In order to produce useful agents, however, it is useful to be able to control the degree of adversariality by specifying rules that an agent must follow. For example, when testing an autonomous vehicle, it is useful to find maximally antagonistic traffic participants that obey traffic rules. We model dynamic constraints as hierarchically ordered rules expressed in Signal T emporal Logic, and show how these can be incorporated into an agent training process. We prove that our agent-centric approach is able to find all dangerous behaviors that can be found by traditional falsification techniques while producing modular and reusable agents. We demonstrate our approach on two case studies from the automotive domain. I NTRODUCTION When developing cyberphysical systems such as autonomous vehicles, drones, or aircraft, it is important to have a robust testing strategy that finds critical bugs before the system is put into production. Falsification techniques exist to find simulations in which the system under test fails to satisfy its target specification. These falsification traces can be generated from a bounded set of inputs.

Model-agnostic meta-learners aim to acquire meta-learned parameters from similar tasks to adapt to novel tasks from the same distribution with few gradient updates. With the flexibility in the choice of models, those frameworks demonstrate appealing performance on a variety of domains such as few-shot image classification and reinforcement learning. However, one important limitation of such frameworks is that they seek a common initialization shared across the entire task distribution, substantially limiting the diversity of the task distributions that they are able to learn from. In this paper, we augment MAML with the capability to identify the mode of tasks sampled from a multimodal task distribution and adapt quickly through gradient updates. Specifically, we propose a multimodal MAML (MMAML) framework, which is able to modulate its meta-learned prior parameters according to the identified mode, allowing more efficient fast adaptation. We evaluate the proposed model on a diverse set of few-shot learning tasks, including regression, image classification, and reinforcement learning. The results not only demonstrate the effectiveness of our model in modulating the meta-learned prior in response to the characteristics of tasks but also show that training on a multimodal distribution can produce an improvement over unimodal training.

The social implications of algorithmic decision-making in sensitive contexts have generated lively debates among multiple stakeholders, suc h as moral and political philosophers, computer scientists, and the public. Yet, the lack of a common language and a conceptual framework for an appropriate bridging of the mor al, technical, and political aspects of the debate prevents the discussion to be as effective a s it can be. Social scientists and psychologists are contributing to this debate by gather ing a wealth of empirical data, yet a philosophical analysis of the social implications of a lgorithmic decision-making remains comparatively impoverished. In attempting to address this lacuna, this paper argues that a hierarchy of different types of explanations for why and how an algorithmic decision outcome is achieved can establish the relevant connection between t he moral and technical aspects of algorithmic decision-making. In particular, I offer a multifaceted conceptual framework for the explanations and the interpretations of algorithmic de cisions, and I claim that this framework can lay the groundwork for a focused discussion among mu ltiple stakeholders about the social implications of algorithmic decision-making, as we ll as AI governance and ethics more generally.

Imitation learning provides a way to automatically construct a controller by mimicking human behavior from data. For safety-critical systems such as autonomous vehicles, it can be problematic to use controllers learned from data because they cannot be guaranteed to be collision-free. Recently, a method has been proposed for learning a multi-mode hybrid automaton cruise controller (MOHA). Besides being accurate, the logical nature of this model makes it suitable for formal verification. In this paper, we demonstrate this capability using the SpaceEx hybrid model checker as follows. After learning, we translate the automaton model into constraints and equations required by SpaceEx. We then verify that a pure MOHA controller is not collision-free. By adding a safety state based on headway in time, a rule that human drivers should follow anyway, we do obtain a provably safe cruise control. Moreover, the safe controller remains more human-like than existing cruise controllers.

Interpretability is an elusive but highly sought-after characteristic of modern machine learning methods. Recent work has focused on interpretability via $\textit{explanations}$, which justify individual model predictions. In this work, we take a step towards reconciling machine explanations with those that humans produce and prefer by taking inspiration from the study of explanation in philosophy, cognitive science, and the social sciences. We identify key aspects in which these human explanations differ from current machine explanations, distill them into a list of desiderata, and formalize them into a framework via the notion of $\textit{weight of evidence}$ from information theory. Finally, we instantiate this framework in two simple applications and show it produces intuitive and comprehensible explanations.

One of the most common mistakes made when performing data analysis is attributing causal meaning to regression coefficients. Formally, a causal effect can only be computed if it is identifiable from a combination of observational data and structural knowledge about the domain under investigation (Pearl, 2000, Ch. 5). Building on the literature of instrumental variables (IVs), a plethora of methods has been developed to identify causal effects in linear systems. Almost invariably, however, the most powerful such methods rely on exponential-time procedures. In this paper, we investigate graphical conditions to allow efficient identification in arbitrary linear structural causal models (SCMs). In particular, we develop a method to efficiently find unconditioned instrumental subsets, which are generalizations of IVs that can be used to tame the complexity of many canonical algorithms found in the literature. Further, we prove that determining whether an effect can be identified with TSID (Weihs et al., 2017), a method more powerful than unconditioned instrumental sets and other efficient identification algorithms, is NP-Complete. Finally, building on the idea of flow constraints, we introduce a new and efficient criterion called Instrumental Cutsets (IC), which is able to solve for parameters missed by all other existing polynomial-time algorithms.

In reinforcement learning (RL), an autonomous agent learns to perform complex tasks by maximizing an exogenous reward signal while interacting with its environment. In real-world applications, test conditions may differ substantially from the training scenario and, therefore, focusing on pure reward maximization during training may lead to poor results at test time. In these cases, it is important to trade-off between performance and robustness while learning a policy. While several results exist for robust, model-based RL, the model-free case has not been widely investigated. In this paper, we cast the robust, model-free RL problem as a multi-objective optimization problem. To quantify the robustness of a policy, we use delay margin and gain margin, two robustness indicators that are common in control theory. We show how these metrics can be estimated from data in the model-free setting. We use multi-objective Bayesian optimization (MOBO) to solve efficiently this expensive-to-evaluate, multi-objective optimization problem. We show the benefits of our robust formulation both in sim-to-real and pure hardware experiments to balance a Furuta pendulum.

We present a novel approach to control design for nonlinear systems, which leverages reinforcement learning techniques to learn a linearizing controller for a physical plant with unknown dynamics. Feedback linearization is a technique from nonlinear control which renders the input-output dynamics of a nonlinear plant \emph{linear} under application of an appropriate feedback controller. Once a linearizing controller has been constructed, desired output trajectories for the nonlinear plant can be tracked using a variety of linear control techniques. A single learned policy then serves to track arbitrary desired reference signals provided by a higher-level planner. We present theoretical results which provide conditions under which the learning problem has a unique solution which exactly linearizes the plant. We demonstrate the performance of our approach on two simulated problems and a physical robotic platform. For the simulated environments, we observe that the learned feedback linearizing policies can achieve arbitrary tracking of reference trajectories for a fully actuated double pendulum and a 14 dimensional quadrotor. In hardware, we demonstrate that our approach significantly improves tracking performance on a 7-DOF Baxter robot after less than two hours of training.

Using neural networks in the reinforcement learning (RL) framework has achieved notable successes. Yet, neural networks tend to forget what they learned in the past, especially when they learn online and fully incrementally, a setting in which the weights are updated after each sample is received and the sample is then discarded. Under this setting, an update can lead to overly global generalization by changing too many weights. The global generalization interferes with what was previously learned and deteriorates performance, a phenomenon known as catastrophic interference. Many previous works use mechanisms such as experience replay (ER) buffers to mitigate interference by performing minibatch updates, ensuring the data distribution is approximately independent-and-identically-distributed (i.i.d.). But using ER would become infeasible in terms of memory as problem complexity increases. Thus, it is crucial to look for more memory-efficient alternatives. Interference can be averted if we replace global updates with more local ones, so only weights responsible for the observed data sample are updated. In this work, we propose the use of dynamic self-organizing map (DSOM) with neural networks to induce such locality in the updates without ER buffers. Our method learns a DSOM to produce a mask to reweigh each hidden unit's output, modulating its degree of use. It prevents interference by replacing global updates with local ones, conditioned on the agent's state. We validate our method on standard RL benchmarks including Mountain Car and Lunar Lander, where existing methods often fail to learn without ER. Empirically, we show that our online and fully incremental method is on par with and in some cases, better than state-of-the-art in terms of final performance and learning speed. We provide visualizations and quantitative measures to show that our method indeed mitigates interference.