Search

Designing reward functions is a core component of reinforcement learning but can be challenging for truly complex behavior. Reinforcement Learning from Human Feedback (RLHF) has been used to alleviate this challenge by replacing a hand-coded reward function with a reward function learned from preferences. However, it can be exceedingly inefficient to learn these rewards as they are often learned tabula rasa. We investigate whether Large Language Models (LLMs) can reduce this query inefficiency by converting an iterative series of human preferences into code representing the rewards. We propose In-Context Preference Learning (ICPL), a method that uses the grounding of an LLM to accelerate learning reward functions from preferences. ICPL takes the environment context and task description, synthesizes a set of reward functions, and then repeatedly updates the reward functions using human rankings of videos of the resultant policies. Using synthetic preferences, we demonstrate that ICPL is orders of magnitude more efficient than RLHF and is even competitive with methods that use ground-truth reward functions instead of preferences. Finally, we perform a series of human preference-learning trials and observe that ICPL extends beyond synthetic settings and can work effectively with humans-in-the-loop. Additional information and videos are provided at https://sites.google.com/view/few-shot-icpl/home.

Multi-agent learning algorithms have been successful at generating superhuman planning in various games but have had limited impact on the design of deployed multi-agent planners. A key bottleneck in applying these techniques to multi-agent planning is that they require billions of steps of experience. To enable the study of multi-agent planning at scale, we present GPUDrive, a GPU-accelerated, multi-agent simulator built on top of the Madrona Game Engine that can generate over a million simulation steps per second. Observation, reward, and dynamics functions are written directly in C++, allowing users to define complex, heterogeneous agent behaviors that are lowered to high-performance CUDA. We show that using GPUDrive we can effectively train reinforcement learning agents over many scenes in the Waymo Open Motion Dataset, yielding highly effective goal-reaching agents in minutes for individual scenes and enabling agents to navigate thousands of scenarios within hours. The code base with pre-trained agents is available at \url{https://github.com/Emerge-Lab/gpudrive}.

A central challenge for autonomous vehicles is coordinating with humans. Therefore, incorporating realistic human agents is essential for scalable training and evaluation of autonomous driving systems in simulation. Simulation agents are typically developed by imitating large-scale, high-quality datasets of human driving. However, pure imitation learning agents empirically have high collision rates when executed in a multi-agent closed-loop setting. To build agents that are realistic and effective in closed-loop settings, we propose Human-Regularized PPO (HR-PPO), a multi-agent algorithm where agents are trained through self-play with a small penalty for deviating from a human reference policy. In contrast to prior work, our approach is RL-first and only uses 30 minutes of imperfect human demonstrations. We evaluate agents in a large set of multi-agent traffic scenes. Results show our HR-PPO agents are highly effective in achieving goals, with a success rate of 93%, an off-road rate of 3.5%, and a collision rate of 3%. At the same time, the agents drive in a human-like manner, as measured by their similarity to existing human driving logs. We also find that HR-PPO agents show considerable improvements on proxy measures for coordination with human driving, particularly in highly interactive scenarios. We open-source our code and trained agents at https://github.com/Emerge-Lab/nocturne_lab and provide demonstrations of agent behaviors at https://sites.google.com/view/driving-partners.

In this article, we explore the technical details of the reinforcement learning (RL) algorithms that were deployed in the largest field test of automated vehicles designed to smooth traffic flow in history as of 2023, uncovering the challenges and breakthroughs that come with developing RL controllers for automated vehicles. We delve into the fundamental concepts behind RL algorithms and their application in the context of self-driving cars, discussing the developmental process from simulation to deployment in detail, from designing simulators to reward function shaping. We present the results in both simulation and deployment, discussing the flow-smoothing benefits of the RL controller. From understanding the basics of Markov decision processes to exploring advanced techniques such as deep RL, our article offers a comprehensive overview and deep dive of the theoretical foundations and practical implementations driving this rapidly evolving field. We also showcase real-world case studies and alternative research projects that highlight the impact of RL controllers in revolutionizing autonomous driving. From tackling complex urban environments to dealing with unpredictable traffic scenarios, these intelligent controllers are pushing the boundaries of what automated vehicles can achieve. Furthermore, we examine the safety considerations and hardware-focused technical details surrounding deployment of RL controllers into automated vehicles. As these algorithms learn and evolve through interactions with the environment, ensuring their behavior aligns with safety standards becomes crucial. We explore the methodologies and frameworks being developed to address these challenges, emphasizing the importance of building reliable control systems for automated vehicles.

The CIRCLES project aims to reduce instabilities in traffic flow, which are naturally occurring phenomena due to human driving behavior. These "phantom jams" or "stop-and-go waves,"are a significant source of wasted energy. Toward this goal, the CIRCLES project designed a control system referred to as the MegaController by the CIRCLES team, that could be deployed in real traffic. Our field experiment leveraged a heterogeneous fleet of 100 longitudinally-controlled vehicles as Lagrangian traffic actuators, each of which ran a controller with the architecture described in this paper. The MegaController is a hierarchical control architecture, which consists of two main layers. The upper layer is called Speed Planner, and is a centralized optimal control algorithm. It assigns speed targets to the vehicles, conveyed through the LTE cellular network. The lower layer is a control layer, running on each vehicle. It performs local actuation by overriding the stock adaptive cruise controller, using the stock on-board sensors. The Speed Planner ingests live data feeds provided by third parties, as well as data from our own control vehicles, and uses both to perform the speed assignment. The architecture of the speed planner allows for modular use of standard control techniques, such as optimal control, model predictive control, kernel methods and others, including Deep RL, model predictive control and explicit controllers. Depending on the vehicle architecture, all onboard sensing data can be accessed by the local controllers, or only some. Control inputs vary across different automakers, with inputs ranging from torque or acceleration requests for some cars, and electronic selection of ACC set points in others. The proposed architecture allows for the combination of all possible settings proposed above. Most configurations were tested throughout the ramp up to the MegaVandertest.

This review outlines the main results which show the dual nature of the chemical bond in diatomic beryllium molecule in the ground $X^1\Sigma_g^+$ state. It has been shown that the beryllium atoms are covalently bound at low-lying vibrational energy levels ({\nu}=0-4), while at higher ones ({\nu}=5-11) they are bound by van der Waals forces near the right turning points. High precision ab initio quantum calculations of Be$_2$ resulted in the development of the modified expanded Morse oscillator potential function which contains all twelve vibrational energy levels [A.V. Mitin, Chem. Phys. Lett. 682, 30 (2017)]. The dual nature of chemical bond in Be$_2$ is evidenced as a sharp corner on the attractive branch of the ground state potential curve. Moreover, it has been found that the Douglas-Kroll-Hess relativistic corrections also show a sharp corner when presented in dependence on the internuclear separation. The difference in energy between the extrapolated and calculated multi-reference configuration interaction energies in dependence on the internuclear separation also exhibits singular point in the same region. The other problems of ab initio quantum calculations of the beryllium dimer are also discussed. Calculated spectrum of vibrational-rotational bound states and new metastable states of the beryllium dimer in the ground state important for laser spectroscopy are presented. The vibration problem was solved for the modified expanded Morse oscillator potential function and for the potential function obtained with Slater-type orbitals [M. Lesiuk et al, Chem. Theory Comput. 15, 2470 (2019)]. The theoretical upper and lower estimates of the spectrum of vibrational-rotational bound states and the spectrum of rotational-vibrational metastable states with complex-valued energy eigenvalues and the scattering length in the beryllium dimer are presented.

A key challenge in training generally-capable agents is the design of training tasks that facilitate broad generalization and robustness to environment variations. This challenge motivates the problem setting of Unsupervised Environment Design (UED), whereby a student agent trains on an adaptive distribution of tasks proposed by a teacher agent. A pioneering approach for UED is PAIRED, which uses reinforcement learning (RL) to train a teacher policy to design tasks from scratch, making it possible to directly generate tasks that are adapted to the agent's current capabilities. Despite its strong theoretical backing, PAIRED suffers from a variety of challenges that hinder its practical performance. Thus, state-of-the-art methods currently rely on curation and mutation rather than generation of new tasks. In this work, we investigate several key shortcomings of PAIRED and propose solutions for each shortcoming. As a result, we make it possible for PAIRED to match or exceed state-of-the-art methods, producing robust agents in several established challenging procedurally-generated environments, including a partially-observed maze navigation task and a continuous-control car racing environment. We believe this work motivates a renewed emphasis on UED methods based on learned models that directly generate challenging environments, potentially unlocking more open-ended RL training and, as a result, more general agents., Comment: CoLLAs 2023 - Oral; Second and third authors contributed equally

Emerging vehicular systems with increasing proportions of automated components present opportunities for optimal control to mitigate congestion and increase efficiency. There has been a recent interest in applying deep reinforcement learning (DRL) to these nonlinear dynamical systems for the automatic design of effective control strategies. Despite conceptual advantages of DRL being model-free, studies typically nonetheless rely on training setups that are painstakingly specialized to specific vehicular systems. This is a key challenge to efficient analysis of diverse vehicular and mobility systems. To this end, this article contributes a streamlined methodology for vehicular microsimulation and discovers high performance control strategies with minimal manual design. A variable-agent, multi-task approach is presented for optimization of vehicular Partially Observed Markov Decision Processes. The methodology is experimentally validated on mixed autonomy traffic systems, where fractions of vehicles are automated; empirical improvement, typically 15-60% over a human driving baseline, is observed in all configurations of six diverse open or closed traffic systems. The study reveals numerous emergent behaviors resembling wave mitigation, traffic signaling, and ramp metering. Finally, the emergent behaviors are analyzed to produce interpretable control strategies, which are validated against the learned control strategies., Comment: 16 pages, 14 figures, IEEE Transactions on Automation Science and Engineering (T-ASE), 2022

We introduce Nocturne, a new 2D driving simulator for investigating multi-agent coordination under partial observability. The focus of Nocturne is to enable research into inference and theory of mind in real-world multi-agent settings without the computational overhead of computer vision and feature extraction from images. Agents in this simulator only observe an obstructed view of the scene, mimicking human visual sensing constraints. Unlike existing benchmarks that are bottlenecked by rendering human-like observations directly using a camera input, Nocturne uses efficient intersection methods to compute a vectorized set of visible features in a C++ back-end, allowing the simulator to run at over 2000 steps-per-second. Using open-source trajectory and map data, we construct a simulator to load and replay arbitrary trajectories and scenes from real-world driving data. Using this environment, we benchmark reinforcement-learning and imitation-learning agents and demonstrate that the agents are quite far from human-level coordination ability and deviate significantly from the expert trajectories.

Society is characterized by the presence of a variety of social norms: collective patterns of sanctioning that can prevent miscoordination and free-riding. Inspired by this, we aim to construct learning dynamics where potentially beneficial social norms can emerge. Since social norms are underpinned by sanctioning, we introduce a training regime where agents can access all sanctioning events but learning is otherwise decentralized. This setting is technologically interesting because sanctioning events may be the only available public signal in decentralized multi-agent systems where reward or policy-sharing is infeasible or undesirable. To achieve collective action in this setting we construct an agent architecture containing a classifier module that categorizes observed behaviors as approved or disapproved, and a motivation to punish in accord with the group. We show that social norms emerge in multi-agent systems containing this agent and investigate the conditions under which this helps them achieve socially beneficial outcomes.

This work presents an integrated framework of: vehicle dynamics models, with a particular attention to instabilities and traffic waves; vehicle energy models, with particular attention to accurate energy values for strongly unsteady driving profiles; and sparse Lagrangian controls via automated vehicles, with a focus on controls that can be executed via existing technology such as adaptive cruise control systems. This framework serves as a key building block in developing control strategies for human-in-the-loop traffic flow smoothing on real highways. In this contribution, we outline the fundamental merits of integrating vehicle dynamics and energy modeling into a single framework, and we demonstrate the energy impact of sparse flow smoothing controllers via simulation results.

Proximal Policy Optimization (PPO) is a ubiquitous on-policy reinforcement learning algorithm but is significantly less utilized than off-policy learning algorithms in multi-agent settings. This is often due to the belief that PPO is significantly less sample efficient than off-policy methods in multi-agent systems. In this work, we carefully study the performance of PPO in cooperative multi-agent settings. We show that PPO-based multi-agent algorithms achieve surprisingly strong performance in four popular multi-agent testbeds: the particle-world environments, the StarCraft multi-agent challenge, Google Research Football, and the Hanabi challenge, with minimal hyperparameter tuning and without any domain-specific algorithmic modifications or architectures. Importantly, compared to competitive off-policy methods, PPO often achieves competitive or superior results in both final returns and sample efficiency. Finally, through ablation studies, we analyze implementation and hyperparameter factors that are critical to PPO's empirical performance, and give concrete practical suggestions regarding these factors. Our results show that when using these practices, simple PPO-based methods can be a strong baseline in cooperative multi-agent reinforcement learning. Source code is released at \url{https://github.com/marlbenchmark/on-policy}., Comment: This paper has been accepted by NeurIPS 2022 Datasets and Benchmarks

A wide range of reinforcement learning (RL) problems - including robustness, transfer learning, unsupervised RL, and emergent complexity - require specifying a distribution of tasks or environments in which a policy will be trained. However, creating a useful distribution of environments is error prone, and takes a significant amount of developer time and effort. We propose Unsupervised Environment Design (UED) as an alternative paradigm, where developers provide environments with unknown parameters, and these parameters are used to automatically produce a distribution over valid, solvable environments. Existing approaches to automatically generating environments suffer from common failure modes: domain randomization cannot generate structure or adapt the difficulty of the environment to the agent's learning progress, and minimax adversarial training leads to worst-case environments that are often unsolvable. To generate structured, solvable environments for our protagonist agent, we introduce a second, antagonist agent that is allied with the environment-generating adversary. The adversary is motivated to generate environments which maximize regret, defined as the difference between the protagonist and antagonist agent's return. We call our technique Protagonist Antagonist Induced Regret Environment Design (PAIRED). Our experiments demonstrate that PAIRED produces a natural curriculum of increasingly complex environments, and PAIRED agents achieve higher zero-shot transfer performance when tested in highly novel environments.

We analyse the origin of the unexpected deep sub-barrier heavy-ion fusion hindrance in 64Ni+100Mo and 28Si+64Ni recations. Our analysis is based on the improved coupled-channels approach, implemented by means of the finite element method. With the aid of the Woods-Saxon potential the experimental cross sections and the S-factors of these reactions are remarkably well reproduced. We found that the account on the non-diagonal matrix elements of the coupling matrix, traditionally neglected in the conventional coupled-channels approaches in setting the left boundary conditions inside the potential pocket, and its minimal value are crucially important for the interpretation experimental data. Within our approach we found a good agreement with the experimental data for the S-factor of the fusion reaction 12C+12C, which has no a pronounced maximum for this system., Comment: 6 pages, 4 figures

We study the ability of autonomous vehicles to improve the throughput of a bottleneck using a fully decentralized control scheme in a mixed autonomy setting. We consider the problem of improving the throughput of a scaled model of the San Francisco-Oakland Bay Bridge: a two-stage bottleneck where four lanes reduce to two and then reduce to one. Although there is extensive work examining variants of bottleneck control in a centralized setting, there is less study of the challenging multi-agent setting where the large number of interacting AVs leads to significant optimization difficulties for reinforcement learning methods. We apply multi-agent reinforcement algorithms to this problem and demonstrate that significant improvements in bottleneck throughput, from 20\% at a 5\% penetration rate to 33\% at a 40\% penetration rate, can be achieved. We compare our results to a hand-designed feedback controller and demonstrate that our results sharply outperform the feedback controller despite extensive tuning. Additionally, we demonstrate that the RL-based controllers adopt a robust strategy that works across penetration rates whereas the feedback controllers degrade immediately upon penetration rate variation. We investigate the feasibility of both action and observation decentralization and demonstrate that effective strategies are possible using purely local sensing. Finally, we open-source our code at https://github.com/eugenevinitsky/decentralized_bottlenecks.

Reinforcement Learning (RL) is an effective tool for controller design but can struggle with issues of robustness, failing catastrophically when the underlying system dynamics are perturbed. The Robust RL formulation tackles this by adding worst-case adversarial noise to the dynamics and constructing the noise distribution as the solution to a zero-sum minimax game. However, existing work on learning solutions to the Robust RL formulation has primarily focused on training a single RL agent against a single adversary. In this work, we demonstrate that using a single adversary does not consistently yield robustness to dynamics variations under standard parametrizations of the adversary; the resulting policy is highly exploitable by new adversaries. We propose a population-based augmentation to the Robust RL formulation in which we randomly initialize a population of adversaries and sample from the population uniformly during training. We empirically validate across robotics benchmarks that the use of an adversarial population results in a more robust policy that also improves out-of-distribution generalization. Finally, we demonstrate that this approach provides comparable robustness and generalization as domain randomization on these benchmarks while avoiding a ubiquitous domain randomization failure mode.

In this article, we demonstrate a zero-shot transfer of an autonomous driving policy from simulation to University of Delaware's scaled smart city with adversarial multi-agent reinforcement learning, in which an adversary attempts to decrease the net reward by perturbing both the inputs and outputs of the autonomous vehicles during training. We train the autonomous vehicles to coordinate with each other while crossing a roundabout in the presence of an adversary in simulation. The adversarial policy successfully reproduces the simulated behavior and incidentally outperforms, in terms of travel time, both a human-driving baseline and adversary-free trained policies. Finally, we demonstrate that the addition of adversarial training considerably improves the performance \eat{stability and robustness} of the policies after transfer to the real world compared to Gaussian noise injection., Comment: 6 pages, 4 figures

Using deep reinforcement learning, we train control policies for autonomous vehicles leading a platoon of vehicles onto a roundabout. Using Flow, a library for deep reinforcement learning in micro-simulators, we train two policies, one policy with noise injected into the state and action space and one without any injected noise. In simulation, the autonomous vehicle learns an emergent metering behavior for both policies in which it slows to allow for smoother merging. We then directly transfer this policy without any tuning to the University of Delaware Scaled Smart City (UDSSC), a 1:25 scale testbed for connected and automated vehicles. We characterize the performance of both policies on the scaled city. We show that the noise-free policy winds up crashing and only occasionally metering. However, the noise-injected policy consistently performs the metering behavior and remains collision-free, suggesting that the noise helps with the zero-shot policy transfer. Additionally, the transferred, noise-injected policy leads to a 5% reduction of average travel time and a reduction of 22% in maximum travel time in the UDSSC. Videos of the controllers can be found at https://sites.google.com/view/iccps-policy-transfer., Comment: To be published at the International Conference on Cyber Physical Systems (ICCPS) 2019. 10 pages, 9 figures

The eigenvalue problem for second-order ordinary differential equation (SOODE) in a finite interval with the boundary conditions of the first, second and third kind is formulated. A computational scheme of the finite element method (FEM) is presented that allows the solution of the eigenvalue problem for a SOODE with the known potential function using the programs ODPEVP and KANTBP 4M that implement FEM in the Fortran and Maple, respectively. Numerical analysis of the solution using the KANTBP 4M program is performed for the SOODE exactly solvable eigenvalue problem. The discrete energy eigenvalues and eigenfunctions are analyzed for vibrational-rotational states of the diatomic beryllium molecule solving the eigenvalue problem for the SOODE numerically with the table-valued potential function approximated by interpolation Lagrange and Hermite polynomials and its asymptotic expansion for large values of the independent variable specified as Fortran function. The efficacy of the programs is demonstrated by the calculations of twelve eigenenergies of vibrational bound states with the required accuracy, in comparison with those known from literature, and the vibrational-rotational spectrum of the diatomic beryllium molecule., Comment: 9 pages, 3 figures, 1 table

We apply a new calculation scheme of a finite element method (FEM) for solving an elliptic boundary-value problem describing a quadrupole vibration collective nuclear model with tetrahedral symmetry. We use of shape functions constructed with interpolation Lagrange polynomials on a triangle finite element grid and compare the FEM results with obtained early by a finite difference method., Comment: 6 pages, 3 figures, 1 table