In autonomous navigation, a planning system reasons about other agents to plan a safe and plausible trajectory. Before planning starts, agents are typically processed with computationally intensive models for recognition, tracking, motion estimation and prediction. With limited computational resources and a large number of agents to process in real time, it becomes important to efficiently rank agents according to their impact on the decision making process. This allows spending more time processing the most important agents. We propose a system to rank agents around an autonomous vehicle (AV) in real time. We automatically generate a ranking data set by running the planner in simulation on real-world logged data, where we can afford to run more accurate and expensive models on all the agents. The causes of various planner actions are logged and used for assigning ground truth importance scores. The generated data set can be used to learn ranking models. In particular, we show the utility of combining learned features, via a convolutional neural network, with engineered features designed to capture domain knowledge. We show the benefits of various design choices experimentally. When tested on real AVs, our system demonstrates the capability of understanding complex driving situations.

Many machine learning applications involve jointly predicting multiple mutually dependent output variables. Learning to search is a family of methods where the complex decision problem is cast into a sequence of decisions via a search space. Although these methods have shown promise both in theory and in practice, implementing them has been burdensomely awkward. In this paper, we show the search space can be defined by an arbitrary imperative program, turning learning to search into a credit assignment compiler. Altogether with the algorithmic improvements for the compiler, we radically reduce the complexity of programming and the running time. We demonstrate the feasibility of our approach on multiple joint prediction tasks. In all cases, we obtain accuracies as high as alternative approaches, at drastically reduced execution and programming time.

Many prediction domains, such as ad placement, recommendation, trajectory prediction, and document summarization, require predicting a set or list of options. Such lists are often evaluated using submodular reward functions that measure both quality and diversity. We propose a simple, efficient, and provably near-optimal approach to optimizing such prediction problems based on no-regret learning. Our method leverages a surprising result from online submodular optimization: a single no-regret online learner can compete with an optimal sequence of predictions. Compared to previous work, which either learn a sequence of classifiers or rely on stronger assumptions such as realizability, we ensure both data-efficiency as well as performance guarantees in the fully agnostic setting. Experiments validate the efficiency and applicability of the approach on a wide range of problems including manipulator trajectory optimization, news recommendation and document summarization.

Autonomous navigation for large Unmanned Aerial Vehicles (UAVs) is fairly straight-forward, as expensive sensors and monitoring devices can be employed. In contrast, obstacle avoidance remains a challenging task for Micro Aerial Vehicles (MAVs) which operate at low altitude in cluttered environments. Unlike large vehicles, MAVs can only carry very light sensors, such as cameras, making autonomous navigation through obstacles much more challenging. In this paper, we describe a system that navigates a small quadrotor helicopter autonomously at low altitude through natural forest environments. Using only a single cheap camera to perceive the environment, we are able to maintain a constant velocity of up to 1.5m/s. Given a small set of human pilot demonstrations, we use recent state-of-the-art imitation learning techniques to train a controller that can avoid trees by adapting the MAVs heading. We demonstrate the performance of our system in a more controlled environment indoors, and in real natural forest environments outdoors.

Nearly every structured prediction problem in computer vision requires approximate inference due to large and complex dependencies among output labels. While graphical models provide a clean separation between modeling and inference, learning these models with approximate inference is not well understood. Furthermore, even if a good model is learned, predictions are often inaccurate due to approximations. In this work, instead of performing inference over a graphical model, we instead consider the inference procedure as a composition of predictors. Specifically, we focus on message-passing algorithms, such as Belief Propagation, and show how they can be viewed as procedures that sequentially predict label distributions at each node over a graph. Given labeled graphs, we can then train the sequence of predictors to output the correct labeling s. The result no longer corresponds to a graphical model but simply defines an inference procedure, with strong theoretical properties, that can be used to classify new graphs. We demonstrate the scalability and efficacy of our approach on 3D point cloud classification and 3D surface estimation from single images.

In sequential decision making under uncertainty, as in many other modeling endeavors, researchers observe a dynamical system and collect data measuring its behavior over time. These data are often used to build models that explain relationships between the measured variables, and are eventually used for planning and control purposes. However, these measurements cannot always be exact, systems can change over time, and discovering these facts or fixing these problems is not always feasible. Therefore it is important to formally describe the degree to which the model can tolerate noise, in order to keep near optimal behavior. The problem of finding tolerance bounds has been the focus of many studies for Markov Decision Processes (MDPs) due to their usefulness in practical applications. In this paper, we consider Partially Observable MDPs (POMDPs), which is a more realistic extension of MDPs with a wider scope of applications. We address two types of perturbations in POMDP model parameters, namely additive and multiplicative, and provide theoretical bounds for the impact of these changes in the value function. Experimental results are provided to illustrate our POMDP perturbation analysis in practice.

Partially Observable Markov Decision Processes (POMDPs) provide a rich mathematical model to handle real-world sequential decision processes but require a known model to be solved by most approaches. However, mainstream POMDP research focuses on the discrete case and this complicates its application to most realistic problems that are naturally modeled using continuous state spaces. In this paper, we consider the problem of optimal control in continuous and partially observable environments when the parameters of the model are unknown. We advocate the use of Gaussian Process Dynamical Models (GPDMs) so that we can learn the model through experience with the environment. Our results on the blimp problem show that the approach can learn good models of the sensors and actuators in order to maximize long-term rewards.

We consider the problem of optimal control in continuous and partially observable environments when the parameters of the model are not known exactly. Partially observable Markov decision processes (POMDPs) provide a rich mathematical model to handle such environments but require a known model to be solved by most approaches. This is a limitation in practice as the exact model parameters are often difficult to specify exactly. We adopt a Bayesian approach where a posterior distribution over the model parameters is maintained and updated through experience with the environment. We propose a particle filter algorithm to maintain the posterior distribution and an online planning algorithm, based on trajectory sampling, to plan the best action to perform under the current posterior. The resulting approach selects control actions which optimally trade-off between 1) exploring the environment to learn the model, 2) identifying the system's state, and 3) exploiting its knowledge in order to maximize long-term rewards. Our preliminary results on a simulated robot navigation problem show that our approach is able to learn good models of the sensors and actuators, and performs as well as if it had the true model.

Partially Observable Markov Decision Processes (POMDPs) provide a rich framework for sequential decision-making under uncertainty in stochastic domains. However, solving a POMDP is often intractable except for small problems due to their complexity. Here, we focus on online approaches that alleviate the computational complexity by computing good local policies at each decision step during the execution. Online algorithms generally consist of a lookahead search to find the best action to execute at each time step in an environment. Our objectives here are to survey the various existing online POMDP methods, analyze their properties and discuss their advantages and disadvantages; and to thoroughly evaluate these online approaches in different environments under various metrics (return, error bound reduction, lower bound improvement). Our experimental results indicate that state-of-the-art online heuristic search methods can handle large POMDP domains efficiently.

Bayesian Reinforcement Learning has generated substantial interest recently, as it provides an elegant solution to the exploration-exploitation trade-off in reinforce- ment learning. However most investigations of Bayesian reinforcement learning to date focus on the standard Markov Decision Processes (MDPs). Our goal is to extend these ideas to the more general Partially Observable MDP (POMDP) framework, where the state is a hidden variable. To address this problem, we in- troduce a new mathematical model, the Bayes-Adaptive POMDP. This new model allows us to (1) improve knowledge of the POMDP domain through interaction with the environment, and (2) plan optimal sequences of actions which can trade- off between improving the model, identifying the state, and gathering reward. We show how the model can be finitely approximated while preserving the value func- tion. We describe approximations for belief tracking and planning in this model. Empirical results on two domains show that the model estimate and agent’s return improve over time, as the agent learns better model estimates.

Planning in partially observable environments remains a challenging problem, despite significant recent advances in offline approximation techniques. A few online methods have also been proposed recently, and proven to be remarkably scalable, but without the theoretical guarantees of their offline counterparts. Thus it seems natural to try to unify offline and online techniques, preserving the theoretical properties of the former, and exploiting the scalability of the latter. In this paper, we provide theoretical guarantees on an anytime algorithm for POMDPs which aims to reduce the error made by approximate offline value iteration algorithms through the use of an efficient online searching procedure. The algorithm uses search heuristics based on an error analysis of lookahead search, to guide the online search towards reachable beliefs with the most potential to reduce error. We provide a general theorem showing that these search heuristics are admissible, and lead to complete and epsilon-optimal algorithms. This is, to the best of our knowledge, the strongest theoretical result available for online POMDP solution methods. We also provide empirical evidence showing that our approach is also practical, and can find (provably) near-optimal solutions in reasonable time.