An Algorithmic Perspective on Imitation Learning

Robots had a great impact on the manufacturing industry ever since the early seventies when companies such as KUKA and ABB started deploying their first industrial robots. These robots merely performed very specific tasks in specific ways within well-defined environments. Still, they proved to be very useful as they could exceed human performance at these tasks. However, in order to enable robots to enter our daily life, they need to become more versatile and need to operate in much less structured environments. This thesis is partly devoted to stretching these limitations by means of learning, namely imitation learning (IL) and inverse reinforcement learning (IRL). Reinforcement learning (RL) is a powerful approach to enable robots to solve a task in an unknown environment. The practitioner describes a desired behavior by specifying a reward function and the robot autonomously interacts with the environment in order to find a control policy that generates high accumulated reward. However, RL is not suitable for teaching new tasks by non-experts because specifying appropriate reward functions can be difficult. Demonstrating the desired behavior is often easier for non-experts. Imitation learning can be used in order to enable the robot to reproduce the demonstrations. However, without explicitly inferring and modeling the intentions of the demonstrations, it can become difficult to solve the task for unseen situations. Inverse reinforcement learning (IRL) therefore aims to infer a reward function from the demonstrations, such that optimizing this reward function yields the desired behavior even for different situations. This thesis introduces a unifying approach to solve the inverse reinforcement learning problem in the same way as the reinforcement learning problem. This is achieved by framing both problems as information projection problems, i.e., we strive to minimize the relative entropy between a probabilistic model of the robot behavior and a given desired distribution. Furthermore, a trust region on the robot behavior is used to stabilize the optimization. For inverse reinforcement learning, the desired distribution is implicitly given by the expert demonstrations. The resulting optimization can be efficiently solved using state-of-the-art reinforcement learning methods. For reinforcement learning, the log-likelihood of the desired distribution is given by the reward function. The resulting optimization problem corresponds to a standard reinforcement learning formulation, except for an additional objective of maximizing the entropy of the robot behavior. This entropy objective adds little overhead to the optimization, but can lead to better exploration and more diversified policies. Trust-region I-projections are not only useful for training robots, but can also be applied to other machine learning problems. I-projections are typically used for variational inference, in order to approximate an intractable distribution by a simpler model. However, the resulting optimization problems are usually optimized based on stochastic gradient descent which often suffers from high variance in the gradient estimates. As trust-region I-projections where shown to be effective for reinforcement learning and inverse reinforcement learning, this thesis also explores their use for variational inference. More specifically, trust-region I-projections are investigated for the problem of approximating an intractable distribution by a Gaussian mixture model (GMM) with an adaptive number of components. GMMs are highly desirable for variational inference because they can yield arbitrary accurate approximations while inference from GMMs is still relatively cheap. In order to make learning the GMM feasible, we derive a lower bound that enables us to decompose the objective function. The optimization can then be performed by iteratively updating individual components using a technique from reinforcement learning. The resulting method is capable of learning approximations of significantly higher quality than existing variational inference methods. Due to the similarity of the underlying optimization problems, the insights gained from our variational inference method are also useful for IL and IRL. Namely, a similar lower bound can be applied also for the I-projection formulation of imitation learning. However, whereas for variational inference the lower bound serves to decompose the objective function, for imitation learning it allows us to provide a reward signal to the robot that does not depend on its behavior. Compared to reward functions that are relative to the current behavior of the robot---which are typical for popular adversarial methods---behavior-independent reward functions have the advantages that we can show convergence even for greedy optimization. Furthermore, behavior-independent reward functions solve the inverse reinforcement learning problem, thereby closing the gap between imitation learning and IRL. However, algorithms derived from our non-adversarial formulation are actually very similar to existing AIL methods, and we can even show that adversarial inverse reinforcement learning (AIRL) is indeed an instance of our formulation. AIRL was derived from an adversarial formulation, and we point out several problems of that derivation. In contrast, we show that AIRL can be straightforwardly derived from out non-adversarial formulation. Furthermore, we demonstrate that the non-adversarial formulation can be also used to derive novel algorithms by presenting a non-adversarial method for offline imitation learning.

Learning from demonstration, or imitation learning, is the process of learning to act in an environment from examples provided by a teacher. Inverse reinforcement learning (IRL) is a specific form of learning from demonstration that attempts to estimate the reward function of a Markov decision process from examples provided by the teacher. The reward function is often considered the most succinct description of a task. In simple applications, the reward function may be known or easily derived from properties of the system and hard coded into the learning process. However, in complex applications, this may not be possible, and it may be easier to learn the reward function by observing the actions of the teacher. This paper provides a comprehensive survey of the literature on IRL. This survey outlines the differences between IRL and two similar methods - apprenticeship learning and inverse optimal control. Further, this survey organizes the IRL literature based on the principal method, describes applications of IRL algorithms, and provides areas of future research.

This chapter provides an overview of the most popular methods of inverse reinforcement learning (IRL) and imitation learning (IL). These methods solve the problem of optimal control in a data-driven way, similarly to reinforcement learning, however with the critical difference that now rewards are not observed. The problem is rather to learn the reward function from the observed behavior of an agent. As behavioral data without rewards is widely available, the problem of learning from such data is certainly very interesting. This chapter provides a moderate-level technical description of the most promising IRL methods, equips the reader with sufficient knowledge to understand and follow the current literature on IRL, and presents examples that use simple simulated environments to evaluate how these methods perform when the “ground-truth” rewards are known. We then present use cases for IRL in quantitative finance which include applications in trading strategy identification, sentiment-based trading, option pricing, inference of portfolio investors, and market modeling.

Reinforcement Learning (RL) algorithms depend on well-defined reward functions for policy optimization. Designing such functions is a complex task, even for domain experts. However, valid task demonstrations can still be collected from moderately skilled individuals. This motivates the use of methods such as Imitation Learning (IL) and Inverse Reinforcement Learning (IRL), where an expert provides demonstrations, allowing the algorithm to infer a policy or reward function that aligns with observed behavior. A common assumption in IRL is that demonstrations come from highly skilled experts. While some studies have explored the impact of suboptimal demonstrators, the influence of intentionally malicious demonstrations remains underexplored. This study introduces an adversarial demonstrator framework that systematically perturbs a subset of demonstrations to manipulate the reward function inferred by an IRL algorithm. Additionally, it quantifies the impact of such adversarial manipulations on the learned policy. Our results show that simply altering 10% of the demonstrations can lead the IRL algorithm to learn a faulty reward function, ultimately degrading the performance of the trained policy by up to three times the effect of adding 10% random trajectories to the result.

Chemical space is routinely explored by machine learning methods to discover interesting molecules, before time-consuming experimental synthesizing is attempted. However, these methods often rely on a graph representation, ignoring 3D information necessary for determining the stability of the molecules. We propose a reinforcement learning (RL) approach for generating molecules in Cartesian coordinates allowing for quantum chemical prediction of the stability. To improve sample-efficiency we learn basic chemical rules from imitation learning (IL) on the GDB-11 database to create an initial model applicable for all stoichiometries. We then deploy multiple copies of the model conditioned on a specific stoichiometry in a RL setting. The models correctly identify low energy molecules in the database and produce novel isomers not found in the training set. Finally, we apply the model to larger molecules to show how RL further refines the IL model in domains far from the training data.

This paper investigates the error accumulation problem in centralized training and decentralized execution (CTDE) policy‐based multi‐agent reinforcement learning (MARL) algorithms, which arises from local observation inaccuracies. To address this issue, we propose a novel MARL algorithm that incorporates imitation learning using local observations. Firstly, by analysing the multi‐agent proximal policy optimization algorithm and examining the problems arising when global states are replaced with local observations, it is proved that insufficient observations can lead to information loss, thereby introducing errors of advantage function, and it is demonstrated that the generalized advantage estimation method accumulates errors during the training process. Then, imitation learning is introduced and a novel training framework that combines reinforcement learning and imitation learning is proposed. During the reinforcement learning phase, an MARL agent trained with global observations acts as an expert. Subsequently, imitation learning is applied to train another agent that mimics the expert's decisions using only local observations. Finally, the effectiveness of this algorithm is verified in some commonly used multi‐agent environments, which demonstrates its superior performance compared to traditional multi‐agent reinforcement learning algorithms.

Autonomous driving technology is constantly developing to a higher level of complex scenes, and there is a growing demand for the utilization of end-to-end data-driven control. However, the end-to-end path tracking process often encounters challenges in learning efficiency and generalization. To address this issue, this paper designs a deep deterministic policy gradient (DDPG)-based reinforcement learning strategy that integrates imitation learning and feedforward exploration in the path following process. In imitation learning, the path tracking control data generated by the model predictive control (MPC) method is used to train an end-to-end steering control model of a deep neural network. Another feedforward exploration behavior is predicted by road curvature and vehicle speed, and adds it and imitation learning to the DDPG reinforcement learning to obtain decision-making experience and action prediction behavior of the path tracking process. In the reinforcement learning process, imitation learning is used to update the pre-training parameters of the actor network, and a feedforward steering technique with random noise is adopted for strategy exploration. In the reward function, a hierarchical progressive reward form and a constrained objective reward function referring to MPC are designed, and the actor-critic network architecture is determined. Finally, the path tracking performance of the designed method is verified by comparing various training results, simulations, and HIL tests. The results show that the designed method can effectively utilize pre-training and feedforward prior experience to obtain optimal path tracking performance of an autonomous vehicle, and has better generalization ability than other methods. This study provides an efficient control scheme for improving the end-to-end control performance of autonomous vehicles.

We address the problem of inverse reinforcement learning in Markov decision processes where the agent is risksensitive. In particular, we model risk-sensitivity in a reinforcement learning framework by making use of models of human decision-making having their origins in behavioral psychology and economics. We propose a gradient-based inverse reinforcement learning algorithm that minimizes a loss function defined on the observed behavior. We demonstrate the performance of the proposed technique on two examples, the first of which is the canonical Grid World example and the second of which is an MDP modeling passengers' decisions regarding ride-sharing. In the latter, we use pricing and travel time data from a ride-sharing company to construct the transition probabilities and rewards of the MDP.

Reinforcement learning (RL) provides a computational model to animal’s autonomous acquisition of behaviors even in an uncertain environment. Inverse reinforcement learning (IRL) is its opposite; given a history of behaviors of an agent, IRL attempts to determine the unknown characteristics, like a reward function, of the agent. Conventional IRL methods usually assume the agent has taken a stationary policy that is optimal in the environment. However, real RL agents do not necessarily take stationary policy, because they are often on the way of adapting to their own environments. Especially when facing an uncertain environment, an intelligent agent should take a mixed (or switching) strategy consisting of an exploitation that is best at the current situation and an exploration to resolve the environmental uncertainty. In this study, we propose a new IRL method that can identify both of a non-stationary policy and a fixed but unknown reward function, based on the behavioral history of a learning agent; in particular, we estimate a change point of the behavior policy from an exploratory one in the agent’s early stage of the learning and an exploitative one in its later learning stage. When applied to a computer simulation during a simple maze task of an agent, our method could identify the change point of the behavior policy and the fixed reward function, only from the agent’s history of behaviors.

The UAV pursuit-evasion strategy based on Deep Deterministic Policy Gradient (DDPG) algorithm is a current research hotspot. However, this algorithm has the defect of low efficiency in sample exploration. To solve this problem, this paper uses the imitation learning (IL) to improve the DDPG exploration strategy. A kind of quasiproportional guidance control law is designed to generate effective learning samples, which are used as the data of the initial experience pool of DDPG algorithm. The UAV pursuit-evasion strategy based on DDPG and imitation learning (IL-DDPG) is proposed, and the algorithm obtains the data from the experience pool for experience playback learning, which improves the exploration efficiency of the algorithm in the initial stage of training and avoids the problem of too many useless exploration in the training process. The simulation results show that the trained pursuit-UAV can flexibly adjust the flight speed and flight attitude to pursuit the evasion-UAV quickly. It also verifies that the improved DDPG algorithm is more effective than the basic DDPG algorithm to improve the training efficiency.

Robot programming by demonstration (PbD) has become a central topic of robotics that spans across general research areas such as human-robot interaction, machine learning, machine vision and motor control. Robot PbD started about 30 years ago, and has grown importantly during the past decade. The rationale for moving from purely preprogrammed robots to very flexible user-based interfaces for training robots to perform a task is three-fold. First and foremost, PbD, also referred to as imitation learning, is a powerful mechanism for reducing the complexity of search spaces for learning. When observing either good or bad examples, one can reduce the search for a possible solution, by either starting the search from the observed good solution (local optima), or conversely, by eliminating from the search space what is known as a bad solution. Imitation learning is, thus, a powerful tool for enhancing and accelerating learning in both animals and artifacts. Second, imitation learning offers an implicit means of training a machine, such that explicit and tedious programming of a task by a human user can be minimized or eliminated (Fig. 59.1). Imitation learning is thus a natural means of interacting with a machine that would be accessible to lay people. Left: A robot learns how to make a chess move (namely moving the queen forward) by generalizing across different demonstrations of the task performed in slightly different situations (different starting positions of the hand). The robot records its jointsʼ trajectories and learns to extract what-to-imitate, i.e. that the task constraints are reduced to a subpart of the motion located in a plane defined by the three chess pieces. Right: The robot reproduces the skill in a new context (for different initial position of the chess piece) by finding an appropriate controller that satisfies both the task constraints and constraints relative to its body limitation (how-to-imitate problem), adapted from [59.1] Third, studying and modeling the coupling of perception and action, which is at the core of imitation learning, helps us to understand the mechanisms by which the self-organization of perception and action could arise during development. The reciprocal interaction of perception and action could explain how competence in motor control can be grounded in rich structure of perceptual variables, and vice versa, how the processes of perception can develop as means to create successful actions. PbD promises were thus multiple. On the one hand, one hoped that it would make learning faster, in contrast to tedious reinforcement learning methods or trials-and-error learning. On the other hand, one expected that the methods, being user-friendly, would enhance the application of robots in human daily environments. Recent progresses in the field, which we review in this chapter, show that the field has made a leap forward during the past decade toward these goals. In addition, we anticipate that these promises may be fulfilled very soon. Section 59.1 presents a brief historical overview of robot Programming by Demonstration (PbD), introducing several issues that will be discussed later in this chapter. Section 59.2 reviews engineering approaches to robot PbD with an emphasis on machine learning approaches that provide the robot with the ability to adapt the learned skill to different situations (Sect. 59.2.1). This section discusses also the different types of representation that one may use to encode a skill and presents incremental learning techniques to refine the skill progressively (Sect. 59.2.4). Section 59.2.3 emphasizes the importance to give the teacher an active role during learning and presents different ways in which the user can convey cues to the robot to help it to improve its learning. Section 59.2.4 discusses how PbD can be jointly used with other learning strategies to overcome some limitations of PbD. Section 59.3 reviews works that take a more biological approach to robot PbD and develops models of either the cognitive or neural processes of imitation learning in primates. Finally, Sect. 59.4 lists various open issues in robot PbD that have yet been little explored by the field. KeywordsHide Markov ModelReinforcement LearningAugmented RealityGaussian Mixture ModelHumanoid RobotThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Model predictive control (MPC) has established itself as the standard advanced process control method. However, solving the resulting optimization problems in real-time can be challenging, especially when uncertainty is explicitly considered in a robust nonlinear predictive control approach. An increasingly popular alternative to avoid the online solution of the resulting optimization problems is to approximate their solution using neural networks. The networks are trained using many solutions of the MPC problem for different system states and therefore this approach is often called imitation learning. Controllers obtained via imitation learning have important drawbacks, since it is difficult to learn behaviors that are not well represented in the data and they must be redesigned from scratch when the control task changes. In this work, we show that these two drawbacks can be mitigated by combining imitation learning and concepts from reinforcement learning. The central idea is to use imitation learning as a very good initialization of a control policy that is iteratively updated using reinforcement learning, taking advantage of the fact that an explicit and differentiable expression of the approximate MPC controller is available. The efficacy of the combination of the two learning paradigms is highlighted via simulations of a semi-batch industrial polymerization reactor.

Generating paraphrases from given sentences involves decoding words step by step from a large vocabulary. To learn a decoder, supervised learning which maximizes the likelihood of tokens always suffers from the exposure bias. Although both reinforcement learning (RL) and imitation learning (IL) have been widely used to alleviate the bias, the lack of direct comparison leads to only a partial image on their benefits. In this work, we present an empirical study on how RL and IL can help boost the performance of generating paraphrases, with the pointer-generator as a base model. Experiments on the benchmark datasets show that (1) imitation learning is constantly better than reinforcement learning; and (2) the pointer-generator models with imitation learning outperform the state-of-the-art methods with a large margin.

Planning and decision making are closely interconnected processes that often occur in tandem, influence and informing each other. Planning usually precedes decision making in the chronological sequence, and it can be viewed as a strategy to make decisions. A comprehensive planning or decision strategy can facilitate effective decisions. Thus, understanding and learning human decision-making strategies has drawn intensive attention from the AI community. For example, applying planning algorithms into reinforcement leaning (RL) can simulate the consequence of different actions and select optimal decisions based on learned models, while inverse reinforcement learning (IRL) learns a reward function and policy from expert demonstration and applies them into new scenarios. Most of these methods work based on learning human decision strategies by using modeling of a Markovian decision-making process (MDP). In this paper, we argue that the property of MDP is not fit for human decision-making processes in the real-world and it is insufficient to capture human decision strategies. To tackle this challenge, we propose a new approach to identify the characteristics of human decision-making processes as a decision map, where the decision strategy is defined by the probability distribution of human decisions that are adaptive to the dynamic changes in the environment. The proposed approach was inspired by imitation learning (IL) but with fundamental differences: (a) Instead of aiming to learn an optimal policy based on expert’s demonstrations, we aimed to estimate the distribution of decisions of any group of people. (b) Instead of modeling the environment by an MDP, we used an ambiguity probability model to consider the uncertainty of each decision. (c) The participant trajectory was obtained by categorizing each decision of a participant to a certain cluster based on the commonness in the distribution of decisions. The result shows a feasible way to capture human long-term decision dependency, which provides a complement to the existing machine learning methods for understanding and learning human decision strategies.

Imitation learning refers to an agent’s ability to mimic a desired behavior by learning from observations. A major challenge facing learning from demonstrations is to represent the demonstrations in a manner that is adequate for learning and efficient for real time decisions. Creating feature representations is especially challenging when extracted from high dimensional visual data. In this paper, we present a method for imitation learning from raw visual data. The proposed method is applied to a popular imitation learning domain that is relevant to a variety of real life applications; namely navigation. To create a training set, a teacher uses an optimal policy to perform a navigation task, and the actions taken are recorded along with visual footage from the first person perspective. Features are automatically extracted and used to learn a policy that mimics the teacher via a deep convolutional neural network. A trained agent can then predict an action to perform based on the scene it finds itself in. This method is generic, and the network is trained without knowledge of the task, targets or environment in which it is acting. Another common challenge in imitation learning is generalizing a policy over unseen situation in training data. To address this challenge, the learned policy is subsequently improved by employing active learning. While the agent is executing a task, it can query the teacher for the correct action to take in situations where it has low confidence. The active samples are added to the training set and used to update the initial policy. The proposed approach is demonstrated on 4 different tasks in a 3D simulated environment. The experiments show that an agent can effectively perform imitation learning from raw visual data for navigation tasks and that active learning can significantly improve the initial policy using a small number of samples. The simulated testbed facilitates reproduction of these results and comparison with other approaches.