In the 1970s, Pong was a very popular video arcade game. It is a 2D video game emulating table tennis, i.e. you got a bat (a rectangle) you can move vertically and try to hit a "ball" (a moving square). If the ball hits the bounding box of the game, it bounces back like a billiard ball. If you miss the ball, the opponent scores. A single-player adaptation Breakout came out later, where the ball had the ability to destroy some blocks on the top of the screen and the bat moved to the bottom of the screen.

We are motivated by the goal of generalist robots that can complete a wide range of tasks across many environments. Critical to this is the robot's ability to acquire some metric of task success or reward, which is necessary for reinforcement learning, planning, or knowing when to ask for help. For a general-purpose robot operating in the real world, this reward function must also be able to generalize broadly across environments, tasks, and objects, while depending only on on-board sensor observations (e.g. RGB images). While deep learning on large and diverse datasets has shown promise as a path towards such generalization in computer vision and natural language, collecting high quality datasets of robotic interaction at scale remains an open challenge. In contrast, "in-the-wild" videos of humans (e.g. YouTube) contain an extensive collection of people doing interesting tasks across a diverse range of settings. In this work, we propose a simple approach, Domain-agnostic Video Discriminator (DVD), that learns multitask reward functions by training a discriminator to classify whether two videos are performing the same task, and can generalize by virtue of learning from a small amount of robot data with a broad dataset of human videos. We find that by leveraging diverse human datasets, this reward function (a) can generalize zero shot to unseen environments, (b) generalize zero shot to unseen tasks, and (c) can be combined with visual model predictive control to solve robotic manipulation tasks on a real WidowX200 robot in an unseen environment from a single human demo.

We present a deep reinforcement learning-based artificial intelligence agent that could provide optimized development plans given a basic description of the reservoir and rock/fluid properties with minimal computational cost. This artificial intelligence agent, comprising of a convolutional neural network, provides a mapping from a given state of the reservoir model, constraints, and economic condition to the optimal decision (drill/do not drill and well location) to be taken in the next stage of the defined sequential field development planning process. The state of the reservoir model is defined using parameters that appear in the governing equations of the two-phase flow. A feedback loop training process referred to as deep reinforcement learning is used to train an artificial intelligence agent with such a capability. The training entails millions of flow simulations with varying reservoir model descriptions (structural, rock and fluid properties), operational constraints, and economic conditions. The parameters that define the reservoir model, operational constraints, and economic conditions are randomly sampled from a defined range of applicability. Several algorithmic treatments are introduced to enhance the training of the artificial intelligence agent. After appropriate training, the artificial intelligence agent provides an optimized field development plan instantly for new scenarios within the defined range of applicability. This approach has advantages over traditional optimization algorithms (e.g., particle swarm optimization, genetic algorithm) that are generally used to find a solution for a specific field development scenario and typically not generalizable to different scenarios.

In this work, we study the problem of energy-efficient computation offloading enabled by edge computing. In the considered scenario, multiple users simultaneously compete for limited radio and edge computing resources to get offloaded tasks processed under a delay constraint, with the possibility of exploiting low power sleep modes at all network nodes. The radio resource allocation takes into account inter- and intra-cell interference, and the duty cycles of the radio and computing equipment have to be jointly optimized to minimize the overall energy consumption. To address this issue, we formulate the underlying problem as a dynamic long-term optimization. Then, based on Lyapunov stochastic optimization tools, we decouple the formulated problem into a CPU scheduling problem and a radio resource allocation problem to be solved in a per-slot basis. Whereas the first one can be optimally and efficiently solved using a fast iterative algorithm, the second one is solved using distributed multi-agent reinforcement learning due to its non-convexity and NP-hardness. The resulting framework achieves up to 96.5% performance of the optimal strategy based on exhaustive search, while drastically reducing complexity. The proposed solution also allows to increase the network's energy efficiency compared to a benchmark heuristic approach.

General equilibrium macroeconomic models are a core tool used by policymakers to understand a nation's economy. They represent the economy as a collection of forward-looking actors whose behaviours combine, possibly with stochastic effects, to determine global variables (such as prices) in a dynamic equilibrium. However, standard semi-analytical techniques for solving these models make it difficult to include the important effects of heterogeneous economic actors. The COVID-19 pandemic has further highlighted the importance of heterogeneity, for example in age and sector of employment, in macroeconomic outcomes and the need for models that can more easily incorporate it. We use techniques from reinforcement learning to solve such models incorporating heterogeneous agents in a way that is simple, extensible, and computationally efficient. We demonstrate the method's accuracy and stability on a toy problem for which there is a known analytical solution, its versatility by solving a general equilibrium problem that includes global stochasticity, and its flexibility by solving a combined macroeconomic and epidemiological model to explore the economic and health implications of a pandemic. The latter successfully captures plausible economic behaviours induced by differential health risks by age.

This paper introduces an adaptive model-free deep reinforcement approach that can recognize and adapt to the diurnal patterns in the ride-sharing environment with car-pooling. Deep Reinforcement Learning (RL) suffers from catastrophic forgetting due to being agnostic to the timescale of changes in the distribution of experiences. Although RL algorithms are guaranteed to converge to optimal policies in Markov decision processes (MDPs), this only holds in the presence of static environments. However, this assumption is very restrictive. In many real-world problems like ride-sharing, traffic control, etc., we are dealing with highly dynamic environments, where RL methods yield only sub-optimal decisions. To mitigate this problem in highly dynamic environments, we (1) adopt an online Dirichlet change point detection (ODCP) algorithm to detect the changes in the distribution of experiences, (2) develop a Deep Q Network (DQN) agent that is capable of recognizing diurnal patterns and making informed dispatching decisions according to the changes in the underlying environment. Rather than fixing patterns by time of week, the proposed approach automatically detects that the MDP has changed, and uses the results of the new model. In addition to the adaptation logic in dispatching, this paper also proposes a dynamic, demand-aware vehicle-passenger matching and route planning framework that dynamically generates optimal routes for each vehicle based on online demand, vehicle capacities, and locations. Evaluation on New York City Taxi public dataset shows the effectiveness of our approach in improving the fleet utilization, where less than 50% of the fleet are utilized to serve the demand of up to 90% of the requests, while maximizing profits and minimizing idle times.

Underactuated systems like sea vessels have degrees of motion that are insufficiently matched by a set of independent actuation forces. In addition, the underlying trajectory-tracking control problems grow in complexity in order to decide the optimal rudder and thrust control signals. This enforces several difficult-to-solve constraints that are associated with the error dynamical equations using classical optimal tracking and adaptive control approaches. An online machine learning mechanism based on integral reinforcement learning is proposed to find a solution for a class of nonlinear tracking problems with partial prior knowledge of the system dynamics. The actuation forces are decided using innovative forms of temporal difference equations relevant to the vessel's surge and angular velocities. The solution is implemented using an online value iteration process which is realized by employing means of the adaptive critics and gradient descent approaches. The adaptive learning mechanism exhibited well-functioning and interactive features in react to different desired reference-tracking scenarios.

In imitation learning from observation IfO, a learning agent seeks to imitate a demonstrating agent using only observations of the demonstrated behavior without access to the control signals generated by the demonstrator. Recent methods based on adversarial imitation learning have led to state-of-the-art performance on IfO problems, but they typically suffer from high sample complexity due to a reliance on data-inefficient, model-free reinforcement learning algorithms. This issue makes them impractical to deploy in real-world settings, where gathering samples can incur high costs in terms of time, energy, and risk. In this work, we hypothesize that we can incorporate ideas from model-based reinforcement learning with adversarial methods for IfO in order to increase the data efficiency of these methods without sacrificing performance. Specifically, we consider time-varying linear Gaussian policies, and propose a method that integrates the linear-quadratic regulator with path integral policy improvement into an existing adversarial IfO framework. The result is a more data-efficient IfO algorithm with better performance, which we show empirically in four simulation domains: using far fewer interactions with the environment, the proposed method exhibits similar or better performance than the existing technique.

Corrective interventions while a robot is learning to automate a task provide an intuitive method for a human supervisor to assist the robot and convey information about desired behavior. However, these interventions can impose significant burden on a human supervisor, as each intervention interrupts other work the human is doing, incurs latency with each context switch between supervisor and autonomous control, and requires time to perform. We present LazyDAgger, which extends the interactive imitation learning (IL) algorithm SafeDAgger to reduce context switches between supervisor and autonomous control. We find that LazyDAgger improves the performance and robustness of the learned policy during both learning and execution while limiting burden on the supervisor. Simulation experiments suggest that LazyDAgger can reduce context switches by an average of 60% over SafeDAgger on 3 continuous control tasks while maintaining state-of-the-art policy performance. In physical fabric manipulation experiments with an ABB YuMi robot, LazyDAgger reduces context switches by 60% while achieving a 60% higher success rate than SafeDAgger at execution time.

Recently many algorithms were devised for reinforcement learning (RL) with function approximation. While they have clear algorithmic distinctions, they also have many implementation differences that are algorithm-agnostic and sometimes subtle. Such mixing of algorithmic novelty and implementation craftsmanship makes rigorous analyses of the sources of performance improvements difficult. In this work, we focus on a series of inference-based actor-critic algorithms -- MPO, AWR, and SAC -- to decouple their algorithmic innovations and implementation decisions. We present unified derivations through a single control-as-inference objective, where we can categorize each algorithm as based on either Expectation-Maximization (EM) or direct Kullback-Leibler (KL) divergence minimization and treat the rest of specifications as implementation details. We performed extensive ablation studies, and identified substantial performance drops whenever implementation details are mismatched for algorithmic choices. These results show which implementation details are co-adapted and co-evolved with algorithms, and which are transferable across algorithms: as examples, we identified that tanh policy and network sizes are highly adapted to algorithmic types, while layer normalization and ELU are critical for MPO's performances but also transfer to noticeable gains in SAC. We hope our work can inspire future work to further demystify sources of performance improvements across multiple algorithms and allow researchers to build on one another's both algorithmic and implementational innovations.