arXiv subjects

Prashant Doshi

Publications and source records attributed to Prashant Doshi.

At least 37 records · Page 2Linked to original sources

arXiv2021

A Hierarchical Bayesian model for Inverse RL in Partially-Controlled Environments

Robots learning from observations in the real world using inverse reinforcement learning (IRL) may encounter objects or agents in the environment, other than the expert, that cause nuisance observations during the demonstration. These confounding elements are typically removed in fully-controlled environments such as virtual simulations or lab settings. When complete removal is impossible the nuisance observations must be filtered out. However, identifying the source of observations when large amounts of observations are made is difficult. To address this, we present a hierarchical Bayesian model that incorporates both the expert's and the confounding elements' observations thereby explicitly modeling the diverse observations a robot may receive. We extend an existing IRL algorithm originally designed to work under partial occlusion of the expert to consider the diverse observations. In a simulated robotic sorting domain containing both occlusion and confounding elements, we demonstrate the model's effectiveness. In particular, our technique outperforms several other comparative methods, second only to having perfect knowledge of the subject's trajectory.

cs.LG↗

arXiv2021

Many Agent Reinforcement Learning Under Partial Observability

Recent renewed interest in multi-agent reinforcement learning (MARL) has generated an impressive array of techniques that leverage deep reinforcement learning, primarily actor-critic architectures, and can be applied to a limited range of settings in terms of observability and communication. However, a continuing limitation of much of this work is the curse of dimensionality when it comes to representations based on joint actions, which grow exponentially with the number of agents. In this paper, we squarely focus on this challenge of scalability. We apply the key insight of action anonymity, which leads to permutation invariance of joint actions, to two recently presented deep MARL algorithms, MADDPG and IA2C, and compare these instantiations to another recent technique that leverages action anonymity, viz., mean-field MARL. We show that our instantiations can learn the optimal behavior in a broader class of agent networks than the mean-field method, using a recently introduced pragmatic domain.

cs.LG↗

arXiv2020

A Survey of Inverse Reinforcement Learning: Challenges, Methods and Progress

Inverse reinforcement learning (IRL) is the problem of inferring the reward function of an agent, given its policy or observed behavior. Analogous to RL, IRL is perceived both as a problem and as a class of methods. By categorically surveying the current literature in IRL, this article serves as a reference for researchers and practitioners of machine learning and beyond to understand the challenges of IRL and select the approaches best suited for the problem on hand. The survey formally introduces the IRL problem along with its central challenges such as the difficulty in performing accurate inference and its generalizability, its sensitivity to prior knowledge, and the disproportionate growth in solution complexity with problem size. The article elaborates how the current methods mitigate these challenges. We further discuss the extensions to traditional IRL methods for handling: inaccurate and incomplete perception, an incomplete model, multiple reward functions, and nonlinear reward functions. This survey concludes the discussion with some broad advances in the research area and currently open research questions.

cs.LG↗

arXiv2020

Cooperative-Competitive Reinforcement Learning with History-Dependent Rewards

Consider a typical organization whose worker agents seek to collectively cooperate for its general betterment. However, each individual agent simultaneously seeks to act to secure a larger chunk than its co-workers of the annual increment in compensation, which usually comes from a {\em fixed} pot. As such, the individual agent in the organization must cooperate and compete. Another feature of many organizations is that a worker receives a bonus, which is often a fraction of previous year's total profit. As such, the agent derives a reward that is also partly dependent on historical performance. How should the individual agent decide to act in this context? Few methods for the mixed cooperative-competitive setting have been presented in recent years, but these are challenged by problem domains whose reward functions do not depend on the current state and action only. Recent deep multi-agent reinforcement learning (MARL) methods using long short-term memory (LSTM) may be used, but these adopt a joint perspective to the interaction or require explicit exchange of information among the agents to promote cooperation, which may not be possible under competition. In this paper, we first show that the agent's decision-making problem can be modeled as an interactive partially observable Markov decision process (I-POMDP) that captures the dynamic of a history-dependent reward. We present an interactive advantage actor-critic method (IA2C$^+$), which combines the independent advantage actor-critic network with a belief filter that maintains a belief distribution over other agents' models. Empirical results show that IA2C$^+$ learns the optimal policy faster and more robustly than several other baselines including one that uses a LSTM, even when attributed models are incorrect.

cs.LG↗

arXiv2020

SA-Net: Deep Neural Network for Robot Trajectory Recognition from RGB-D Streams

Learning from demonstration (LfD) and imitation learning offer new paradigms for transferring task behavior to robots. A class of methods that enable such online learning require the robot to observe the task being performed and decompose the sensed streaming data into sequences of state-action pairs, which are then input to the methods. Thus, recognizing the state-action pairs correctly and quickly in sensed data is a crucial prerequisite for these methods. We present SA-Net a deep neural network architecture that recognizes state-action pairs from RGB-D data streams. SA-Net performed well in two diverse robotic applications of LfD -- one involving mobile ground robots and another involving a robotic manipulator -- which demonstrates that the architecture generalizes well to differing contexts. Comprehensive evaluations including deployment on a physical robot show that \sanet{} significantly improves on the accuracy of the previous method that utilizes traditional image processing and segmentation.

cs.RO↗

arXiv2020

Active Deception using Factored Interactive POMDPs to Recognize Cyber Attacker's Intent

This paper presents an intelligent and adaptive agent that employs deception to recognize a cyber adversary's intent. Unlike previous approaches to cyber deception, which mainly focus on delaying or confusing the attackers, we focus on engaging with them to learn their intent. We model cyber deception as a sequential decision-making problem in a two-agent context. We introduce factored finitely nested interactive POMDPs (I-POMDPx) and use this framework to model the problem with multiple attacker types. Our approach models cyber attacks on a single honeypot host across multiple phases from the attacker's initial entry to reaching its adversarial objective. The defending I-POMDPx-based agent uses decoys to engage with the attacker at multiple phases to form increasingly accurate predictions of the attacker's behavior and intent. The use of I-POMDPs also enables us to model the adversary's mental state and investigate how deception affects their beliefs. Our experiments in both simulation and on a real host show that the I-POMDPx-based agent performs significantly better at intent recognition than commonly used deception strategies on honeypots.

cs.MA↗

arXiv2020

Recurrent Sum-Product-Max Networks for Decision Making in Perfectly-Observed Environments

Recent investigations into sum-product-max networks (SPMN) that generalize sum-product networks (SPN) offer a data-driven alternative for decision making, which has predominantly relied on handcrafted models. SPMNs computationally represent a probabilistic decision-making problem whose solution scales linearly in the size of the network. However, SPMNs are not well suited for sequential decision making over multiple time steps. In this paper, we present recurrent SPMNs (RSPMN) that learn from and model decision-making data over time. RSPMNs utilize a template network that is unfolded as needed depending on the length of the data sequence. This is significant as RSPMNs not only inherit the benefits of SPMNs in being data driven and mostly tractable, they are also well suited for sequential problems. We establish conditions on the template network, which guarantee that the resulting SPMN is valid, and present a structure learning algorithm to learn a sound template network. We demonstrate that the RSPMNs learned on a testbed of sequential decision-making data sets generate MEUs and policies that are close to the optimal on perfectly-observed domains. They easily improve on a recent batch-constrained reinforcement learning method, which is important because RSPMNs offer a new model-based approach to offline reinforcement learning.

cs.AI↗

arXiv2020

Maximum Entropy Multi-Task Inverse RL

Multi-task IRL allows for the possibility that the expert could be switching between multiple ways of solving the same problem, or interleaving demonstrations of multiple tasks. The learner aims to learn the multiple reward functions that guide these ways of solving the problem. We present a new method for multi-task IRL that generalizes the well-known maximum entropy approach to IRL by combining it with the Dirichlet process based clustering of the observed input. This yields a single nonlinear optimization problem, called MaxEnt Multi-task IRL, which can be solved using the Lagrangian relaxation and gradient descent methods. We evaluate MaxEnt Multi-task IRL in simulation on the robotic task of sorting onions on a processing line where the expert utilizes multiple ways of detecting and removing blemished onions. The method is able to learn the underlying reward functions to a high level of accuracy and it improves on the previous approaches to multi-task IRL.

cs.LG↗

arXiv2020

GrAALF:Supporting Graphical Analysis of Audit Logs for Forensics

System-level audit logs often play a critical role in computer forensics. They capture low-level interactions between programs and users in much detail, making them a rich source of insight and provenance on malicious user activity. However, using these logs to discover and understand malicious activities when a typical computer generates more than 2.5 million system events hourly is both compute and time-intensive. We introduce a graphical system called GrAALF for efficiently loading, storing, processing, querying, and displaying system events to support computer forensics. In comparison to other related systems such as AIQL [13] and SAQL [12], GrAALF offers the flexibility of multiple backend storage solutions, easy-to-use and intuitive querying of logs, and the ability to trace back longer sequences of system events in (near) real-time to help identify and isolate attacks. Equally important, both AIQL and SAQL are not available for public use, whereas GrAALF is open-source. GrAALF offers the choice of compactly storing the logs in main memory, in a relational database system, in a hybrid main memory-database system, and a graph-based database. We compare the responsiveness of each of these options, using multiple huge system-call log files. Next, in multiple real-world attack scenarios, we demonstrate the efficacy and usefulness of GrAALF in identifying the attack and discovering its provenance. Consequently, GrAALF offers a robust solution for analysis of audit logs to support computer forensics.

cs.CR↗

arXiv2019

Scalable Decision-Theoretic Planning in Open and Typed Multiagent Systems

In open agent systems, the set of agents that are cooperating or competing changes over time and in ways that are nontrivial to predict. For example, if collaborative robots were tasked with fighting wildfires, they may run out of suppressants and be temporarily unavailable to assist their peers. We consider the problem of planning in these contexts with the additional challenges that the agents are unable to communicate with each other and that there are many of them. Because an agent's optimal action depends on the actions of others, each agent must not only predict the actions of its peers, but, before that, reason whether they are even present to perform an action. Addressing openness thus requires agents to model each other's presence, which becomes computationally intractable with high numbers of agents. We present a novel, principled, and scalable method in this context that enables an agent to reason about others' presence in its shared environment and their actions. Our method extrapolates models of a few peers to the overall behavior of the many-agent system, and combines it with a generalization of Monte Carlo tree search to perform individual agent reasoning in many-agent open environments. Theoretical analyses establish the number of agents to model in order to achieve acceptable worst case bounds on extrapolation error, as well as regret bounds on the agent's utility from modeling only some neighbors. Simulations of multiagent wildfire suppression problems demonstrate our approach's efficacy compared with alternative baselines.

cs.MA↗

arXiv2018

Reinforcement Learning for Heterogeneous Teams with PALO Bounds

We introduce reinforcement learning for heterogeneous teams in which rewards for an agent are additively factored into local costs, stimuli unique to each agent, and global rewards, those shared by all agents in the domain. Motivating domains include coordination of varied robotic platforms, which incur different costs for the same action, but share an overall goal. We present two templates for learning in this setting with factored rewards: a generalization of Perkins' Monte Carlo exploring starts for POMDPs to canonical MPOMDPs, with a single policy mapping joint observations of all agents to joint actions (MCES-MP); and another with each agent individually mapping joint observations to their own action (MCES-FMP). We use probably approximately local optimal (PALO) bounds to analyze sample complexity, instantiating these templates to PALO learning. We promote sample efficiency by including a policy space pruning technique, and evaluate the approaches on three domains of heterogeneous agents demonstrating that MCES-FMP yields improved policies in less samples compared to MCES-MP and a previous benchmark.

cs.LG↗

arXiv2018

A Framework and Method for Online Inverse Reinforcement Learning

Inverse reinforcement learning (IRL) is the problem of learning the preferences of an agent from the observations of its behavior on a task. While this problem has been well investigated, the related problem of {\em online} IRL---where the observations are incrementally accrued, yet the demands of the application often prohibit a full rerun of an IRL method---has received relatively less attention. We introduce the first formal framework for online IRL, called incremental IRL (I2RL), and a new method that advances maximum entropy IRL with hidden variables, to this setting. Our formal analysis shows that the new method has a monotonically improving performance with more demonstration data, as well as probabilistically bounded error, both under full and partial observability. Experiments in a simulated robotic application of penetrating a continuous patrol under occlusion shows the relatively improved performance and speed up of the new method and validates the utility of online IRL.

cs.LG↗

arXiv2017

Inverse Reinforcement Learning Under Noisy Observations

We consider the problem of performing inverse reinforcement learning when the trajectory of the expert is not perfectly observed by the learner. Instead, a noisy continuous-time observation of the trajectory is provided to the learner. This problem exhibits wide-ranging applications and the specific application we consider here is the scenario in which the learner seeks to penetrate a perimeter patrolled by a robot. The learner's field of view is limited due to which it cannot observe the patroller's complete trajectory. Instead, we allow the learner to listen to the expert's movement sound, which it can also use to estimate the expert's state and action using an observation model. We treat the expert's state and action as hidden data and present an algorithm based on expectation maximization and maximum entropy principle to solve the non-linear, non-convex problem. Related work considers discrete-time observations and an observation model that does not include actions. In contrast, our technique takes expectations over both state and action of the expert, enabling learning even in the presence of extreme noise and broader applications.

cs.RO↗

arXiv2017

Freeway Merging in Congested Traffic based on Multipolicy Decision Making with Passive Actor Critic

Freeway merging in congested traffic is a significant challenge toward fully automated driving. Merging vehicles need to decide not only how to merge into a spot, but also where to merge. We present a method for the freeway merging based on multi-policy decision making with a reinforcement learning method called {\em passive actor-critic} (pAC), which learns with less knowledge of the system and without active exploration. The method selects a merging spot candidate by using the state value learned with pAC. We evaluate our method using real traffic data. Our experiments show that pAC achieves 92\% success rate to merge into a freeway, which is comparable to human decision making.

cs.AI↗

arXiv2017

Actor-Critic for Linearly-Solvable Continuous MDP with Partially Known Dynamics

In many robotic applications, some aspects of the system dynamics can be modeled accurately while others are difficult to obtain or model. We present a novel reinforcement learning (RL) method for continuous state and action spaces that learns with partial knowledge of the system and without active exploration. It solves linearly-solvable Markov decision processes (L-MDPs), which are well suited for continuous state and action spaces, based on an actor-critic architecture. Compared to previous RL methods for L-MDPs and path integral methods which are model based, the actor-critic learning does not need a model of the uncontrolled dynamics and, importantly, transition noise levels; however, it requires knowing the control dynamics for the problem. We evaluate our method on two synthetic test problems, and one real-world problem in simulation and using real traffic data. Our experiments demonstrate improved learning and policy performance.

cs.AI↗

arXiv2016

Dynamic Sum Product Networks for Tractable Inference on Sequence Data (Extended Version)

Sum-Product Networks (SPN) have recently emerged as a new class of tractable probabilistic graphical models. Unlike Bayesian networks and Markov networks where inference may be exponential in the size of the network, inference in SPNs is in time linear in the size of the network. Since SPNs represent distributions over a fixed set of variables only, we propose dynamic sum product networks (DSPNs) as a generalization of SPNs for sequence data of varying length. A DSPN consists of a template network that is repeated as many times as needed to model data sequences of any length. We present a local search technique to learn the structure of the template network. In contrast to dynamic Bayesian networks for which inference is generally exponential in the number of variables per time slice, DSPNs inherit the linear inference complexity of SPNs. We demonstrate the advantages of DSPNs over DBNs and other models on several datasets of sequence data.

cs.LG↗

arXiv2015

Individual Planning in Agent Populations: Exploiting Anonymity and Frame-Action Hypergraphs

Interactive partially observable Markov decision processes (I-POMDP) provide a formal framework for planning for a self-interested agent in multiagent settings. An agent operating in a multiagent environment must deliberate about the actions that other agents may take and the effect these actions have on the environment and the rewards it receives. Traditional I-POMDPs model this dependence on the actions of other agents using joint action and model spaces. Therefore, the solution complexity grows exponentially with the number of agents thereby complicating scalability. In this paper, we model and extend anonymity and context-specific independence -- problem structures often present in agent populations -- for computational gain. We empirically demonstrate the efficiency from exploiting these problem structures by solving a new multiagent problem involving more than 1,000 agents.

cs.MA↗

arXiv2014

Team Behavior in Interactive Dynamic Influence Diagrams with Applications to Ad Hoc Teams

Planning for ad hoc teamwork is challenging because it involves agents collaborating without any prior coordination or communication. The focus is on principled methods for a single agent to cooperate with others. This motivates investigating the ad hoc teamwork problem in the context of individual decision making frameworks. However, individual decision making in multiagent settings faces the task of having to reason about other agents' actions, which in turn involves reasoning about others. An established approximation that operationalizes this approach is to bound the infinite nesting from below by introducing level 0 models. We show that a consequence of the finitely-nested modeling is that we may not obtain optimal team solutions in cooperative settings. We address this limitation by including models at level 0 whose solutions involve learning. We demonstrate that the learning integrated into planning in the context of interactive dynamic influence diagrams facilitates optimal team behavior, and is applicable to ad hoc teamwork.

cs.MA↗