The deployment of pre-trained perception models in novel environments often leads to performance degradation due to distributional shifts. Although recent artificial intelligence approaches for metacognition use logical rules to characterize and filter model errors, improving precision often comes at the cost of reduced recall. This paper addresses the hypothesis that leveraging multiple pre-trained models can mitigate this recall reduction. We formulate the challenge of identifying and managing conflicting predictions from various models as a consistency-based abduction problem, building on the idea of abductive learning (ABL) but applying it to test-time instead of training. The input predictions and the learned error detection rules derived from each model are encoded in a logic program. We then seek an abductive explanation—a subset of model predictions—that maximizes prediction coverage while ensuring the rate of logical inconsistencies (derived from domain constraints) remains below a specified threshold. We propose two algorithms for this knowledge representation task: an exact method based on Integer Programming (IP) and an efficient Heuristic Search (HS). Through extensive experiments on a simulated aerial imagery dataset featuring controlled, complex distributional shifts, we demonstrate that our abduction-based framework outperforms individual models and standard ensemble baselines, achieving, for instance, average relative improvements of approximately 13.6% in F1-score and 16.6% in accuracy across 15 diverse test datasets when compared to the best individual model. Our results validate the use of consistency-based abduction as an effective mechanism to robustly integrate knowledge from multiple imperfect models in challenging, novel scenarios.

Understanding the inner representation of a neural network helps users improve models. Concept-based methods have become a popular choice for explaining deep neural networks post-hoc because, unlike most other explainable AI techniques, they can be used to test high-level visual "concepts" that are not directly related to feature attributes. For instance, the concept of "stripes" is important to classify an image as a zebra. Concept-based explanation methods, however, require practitioners to guess and manually collect multiple candidate concept image sets, making the process labor-intensive and prone to overlooking important concepts. Addressing this limitation, in this paper, we frame concept image set creation as an image generation problem. However, since naively using a standard generative model does not result in meaningful concepts, we devise a reinforcement learning-based preference optimization (RLPO) algorithm that fine-tunes a vision-language generative model from approximate textual descriptions of concepts. Through a series of experiments, we demonstrate our method’s ability to efficiently and reliably articulate diverse concepts that are otherwise challenging to craft manually.

Black box neural networks are an indispensable part of modern robots. Nevertheless, deploying such high-stakes systems in real-world scenarios poses significant challenges when the stakeholders, such as engineers and legislative bodies, lack insights into the neural networks’ decision-making process. Presently, explainable AI is primarily tailored to natural language processing and computer vision, falling short in two critical aspects when applied in robots: grounding in decision-making tasks and the ability to assess trustworthiness of their explanations. In this paper, we introduce a trustworthy explainable robotics technique based on human-interpretable, high-level concepts that attribute to the decisions made by the neural network. Our proposed technique provides explanations with associated uncertainty scores for the explanation by matching neural network’s activations with human-interpretable visualizations. To validate our approach, we conducted a series of experiments with various simulated and real-world robot decision-making models, demonstrating the effectiveness of the proposed approach as a post-hoc, human-friendly robot diagnostic tool. Code: https://github.com/aditya-taparia/BaTCAVe

Vision-Language Models (VLMs) are increasingly pivotal for generalist robot manipulation, enabling tasks such as physical reasoning, policy generation, and failure detection. However, their proficiency in these high-level applications often assumes a deep understanding of low-level physical prerequisites, a capability that is largely unverified. To perform actions reliably, robots must comprehend intrinsic object properties (e.g., material, weight), action affordances (e.g., graspable, stackable), and physical constraints (e.g., stability, reachability, or an object's state like being closed). Despite their ubiquitous use in manipulation, we argue that off-the-shelf VLMs may lack this granular, physically-grounded understanding, as these specific prerequisites are often overlooked during training. Addressing this critical gap, we introduce PAC Bench, a comprehensive benchmark designed to systematically evaluate VLMs on their understanding of these core Properties, Affordances, and Constraints (PAC) from a task executability perspective. PAC Bench features a diverse dataset with more than 30,000 annotations, comprising 673 real-world images (115 object classes, 15 property types, 1–3 affordances defined per object class), 100 real-world humanoid view scenarios, and 120 unique simulated constraint scenarios across four tasks. Our evaluations reveal significant gaps in the ability of VLMs to grasp fundamental physical concepts, underscoring their current limitations for reliable robot manipulation and pointing to key areas that require targeted research. PAC Bench also serves as a standardized benchmark for rigorously evaluating the physical reasoning capabilities of VLMs guiding the development of more robust and physically grounded models for robot manipulation.

In large deep neural networks that seem to perform surprisingly well on many tasks, we also observe a few failures related to accuracy, social biases, and alignment with human values, among others. Therefore, before deploying these models, it is crucial to characterize this failure landscape for engineers to debug and legislative bodies to audit models. Nevertheless, it is infeasible to exhaustively test for all possible combinations of factors that could lead to a model's failure. In this paper, we introduce a post-hoc method that utilizes *deep reinforcement learning* to explore and construct the landscape of failure modes in pre-trained discriminative and generative models. With the aid of limited human feedback, we then demonstrate how to restructure the failure landscape to be more desirable by moving away from the discovered failure modes. We empirically show the effectiveness of the proposed method across common Computer Vision, Natural Language Processing, and Vision-Language tasks.

We generalize the derivation of model predictive path integral control (MPPI) to allow for a single joint distribution across controls in the control sequence. This reformation allows for the implementation of adaptive importance sampling (AIS) algorithms into the original importance sampling step while still maintaining the benefits of MPPI such as working with arbitrary system dynamics and cost functions. The benefit of optimizing the proposal distribution by integrating AIS at each control step is demonstrated in simulated environments including controlling multiple cars around a track. The new algorithm is more sample efficient than MPPI, achieving better performance with fewer samples. This performance disparity grows as the dimension of the action space increases. Results from simulations suggest the new algorithm can be used as an anytime algorithm, increasing the value of control at each iteration versus relying on a large set of samples. Repository—https://github.com/sisl/MPOPIS

An open problem in autonomous vehicle safety validation is building reliable models of human driving behavior in simulation. This work presents an approach to learn neural driving policies from real world driving demonstration data. We model human driving as a sequential decision making problem that is characterized by non-linearity and stochasticity, and unknown underlying cost functions. Imitation learning is an approach for generating intelligent behavior when the cost function is unknown or difficult to specify. Building upon work in inverse reinforcement learning (IRL), Generative Adversarial Imitation Learning (GAIL) aims to provide effective imitation even for problems with large or continuous state and action spaces, such as modeling human driving. This article describes the use of GAIL for learning-based driver modeling. Because driver modeling is inherently a multi-agent problem, where the interaction between agents needs to be modeled, this paper describes a parameter-sharing extension of GAIL called PS-GAIL to tackle multi-agent driver modeling. In addition, GAIL is domain agnostic, making it difficult to encode specific knowledge relevant to driving in the learning process. This paper describes Reward Augmented Imitation Learning (RAIL), which modifies the reward signal to provide domain-specific knowledge to the agent. Finally, human demonstrations are dependent upon latent factors that may not be captured by GAIL. This paper describes Burn-InfoGAIL, which allows for disentanglement of latent variability in demonstrations. Imitation learning experiments are performed using NGSIM, a real-world highway driving dataset. Experiments show that these modifications to GAIL can successfully model highway driving behavior, accurately replicating human demonstrations and generating realistic, emergent behavior in the traffic flow arising from the interaction between driving agents.

Although there has been tremendous progress in autonomous driving, navigating environments and predicting the behavior of other drivers in the presence of occlusions remains challenging. Cities have started investing in infrastructure sensors that could provide information about occluded spaces. We propose a framework that integrates infrastructure-to-vehicle communication in autonomous vehicle decision making, improving operational safety and mobility in challenging environments. By framing the problem as a partially observable Markov decision process in which querying an infrastructure sensor is a data-gathering action, we reduce the computational complexity associated with sensor processing while maintaining equivalent performance compared to an omniscient actor and demonstrate the value of infrastructure communication through a series of experiments.

Autonomous vehicles (AVs) rely on environment perception and behavior prediction to reason about agents in their surroundings. These perception systems must be robust to adverse weather such as rain, fog, and snow. However, validation of these systems is challenging due to their complexity and dependence on observation histories. This paper presents a method for characterizing failures of LiDAR-based perception systems for AVs in adverse weather conditions. We develop a methodology based in reinforcement learning to find likely failures in object tracking and trajectory prediction due to sequences of disturbances. We apply disturbances using a physics-based data augmentation technique for simulating LiDAR point clouds in adverse weather conditions. Experiments performed across a wide range of driving scenarios from a real-world driving dataset show that our proposed approach finds high likelihood failures with smaller input disturbances compared to baselines while remaining computationally tractable. Identified failures can inform future development of robust perception systems for AVs.

We present a method for autonomous exploration of large-scale unknown environments under mission time con-straints. We start by proposing the Frontloaded Information Gain Orienteering Problem (FIG-OP) - a generalization of the traditional orienteering problem where the assumption of a reliable environmental model no longer holds. The FIG-OP ad-dresses model uncertainty by frontloading expected information gain through the addition of a greedy incentive, effectively expe-diting the moment in which new area is uncovered. In order to reason across multi-kilometer environments, we solve FIG-OP over an information-efficient world representation, constructed through the aggregation of information from a topological and metric map. Our method was extensively tested and field-hardened across various complex environments, ranging from subway systems to mines. In comparative simulations, we observe that the FIG-OP solution exhibits improved coverage efficiency over solutions generated by greedy and traditional orienteering-based approaches (i.e. severe and minimal model uncertainty assumptions, respectively).

Deep reinforcement learning (RL) is a promising approach to solving complex robotics problems. However, the process of learning through trial-and-error interactions is often highly time-consuming, despite recent advancements in RL algorithms. Additionally, the success of RL is critically dependent on how well the reward-shaping function suits the task, which is also time-consuming to design. As agents trained on a variety of robotics problems continue to proliferate, the ability to reuse their valuable learning for new domains becomes increasingly significant. In this paper, we propose a post-hoc technique for policy fusion using Optimal Transport theory as a robust means of consolidating the knowledge of multiple agents that have been trained on distinct scenarios. We further demonstrate that this provides an improved weights initialisation of the neural network policy for learning new tasks, requiring less time and computational resources than either retraining the parent policies or training a new policy from scratch. Ultimately, our results on diverse agents commonly used in deep RL show that specialised knowledge can be unified into a “Renaissance agent”, allowing for quicker learning of new skills.

Autonomous vehicles need to model the behavior of surrounding human driven vehicles to be safe and efficient traffic participants. Existing approaches to modeling human driving behavior have relied on both data-driven and rule-based methods. While data-driven models are more expressive, rule-based models are interpretable, which is an important requirement for safety-critical domains like driving. However, rule-based models are not sufficiently representative of data, and data-driven models are yet unable to generate realistic traffic simulation due to unrealistic driving behavior such as collisions. In this paper, we propose a methodology that combines rule-based modeling with data-driven learning. While the rules are governed by interpretable parameters of the driver model, these parameters are learned online from driving demonstration data using particle filtering. We perform driver modeling experiments on the task of highway driving and merging using data from three real-world driving demonstration datasets. Our results show that driver models based on our hybrid rule-based and data-driven approach can accurately capture real-world driving behavior. Further, we assess the realism of the driving behavior generated by our model by having humans perform a “driving Turing test,” where they are asked to distinguish between videos of real driving and those generated using our driver models.

Predicting the future occupancy state of an environment is important to enable informed decisions for autonomous vehicles. Common challenges in occupancy prediction include vanishing dynamic objects and blurred predictions, especially for long prediction horizons. In this work, we propose a double-prong neural network architecture to predict the spatiotemporal evolution of the occupancy state. One prong is dedicated to predicting how the static environment will be observed by the moving ego vehicle. The other prong predicts how the dynamic objects in the environment will move. Experiments conducted on the real-world Waymo Open Dataset indicate that the fused output of the two prongs is capable of retaining dynamic objects and reducing blurriness in the predictions for longer time horizons than baseline models.

Future urban transportation concepts include a mixture of ground and air vehicles with varying degrees of autonomy in a congested environment. In such dynamic environments, occupancy maps alone are not sufficient for safe path planning. Safe and efficient transportation requires reasoning about the 3D flow of traffic and properly modeling uncertainty. Several different approaches can be taken for developing 3D velocity maps. This paper explores a Bayesian approach that captures our uncertainty in the map given training data. The approach involves projecting spatial coordinates into a high-dimensional feature space and then applying Bayesian linear regression to make predictions and quantify uncertainty in our estimates. On a collection of air and ground datasets, we demonstrate that this approach is effective and more scalable than several alternative approaches.

Many applications of generative models rely on the marginalization of their high-dimensional output probability distributions. Normalization functions that yield sparse probability distributions can make exact marginalization more computationally tractable. However, sparse normalization functions usually require alternative loss functions for training since the log-likelihood is undefined for sparse probability distributions. Furthermore, many sparse normalization functions often collapse the multimodality of distributions. In this work, we present ev-softmax, a sparse normalization function that preserves the multimodality of probability distributions. We derive its properties, including its gradient in closed-form, and introduce a continuous family of approximations to ev-softmax that have full support and can be trained with probabilistic loss functions such as negative log-likelihood and Kullback-Leibler divergence. We evaluate our method on a variety of generative models, including variational autoencoders and auto-regressive architectures. Our method outperforms existing dense and sparse normalization techniques in distributional accuracy. We demonstrate that ev-softmax successfully reduces the dimensionality of probability distributions while maintaining multimodality.

Discrete latent spaces in variational autoencoders have been shown to effectively capture the data distribution for many real-world problems such as natural language understanding, human intent prediction, and visual scene representation. However, discrete latent spaces need to be sufficiently large to capture the complexities of real-world data, rendering downstream tasks computationally challenging. For instance, performing motion planning in a high-dimensional latent representation of the environment could be intractable. We consider the problem of sparsifying the discrete latent space of a trained conditional variational autoencoder, while preserving its learned multimodality. As a post hoc latent space reduction technique, we use evidential theory to identify the latent classes that receive direct evidence from a particular input condition and filter out those that do not. Experiments on diverse tasks, such as image generation and human behavior prediction, demonstrate the effectiveness of our proposed technique at reducing the discrete latent sample space size of a model while maintaining its learned multimodality.

Kernel methods have been successfully used in various domains to model nonlinear patterns. However, the structure of the kernels is typically handcrafted for each dataset based on the experience of the data analyst. In this paper, we present a novel technique to learn kernels that best fit the data. We exploit the measure-theoretic view of a shift-invariant kernel given by the Bochner’s theorem, and automatically learn the measure in terms of a parameterized quantile function. This flexible black box quantile function, evaluated on Quasi-Monte Carlo samples, builds up quasi-random Fourier feature maps that can approximate arbitrary kernels. The proposed method is not only general enough to be used in any kernel machine, but can also be combined with other kernel design techniques. We learn expressive kernels on a variety of datasets, verifying the methods ability to automatically discover complex patterns without being guided by human expert knowledge.

This paper addresses the problem of learning instantaneous occupancy levels of dynamic environments and predicting future occupancy levels. Due to the complexity of most real environments, such as urban streets or crowded areas, the efficient and robust incorporation of temporal dependencies into otherwise static occupancy models remains a challenge. We propose a method to capture the uncertainty of moving objects and incorporate this uncertainty information into a continuous occupancy map represented in a rich high-dimensional feature space. This data-efficient model not only allows us to learn the occupancy states incrementally, but also makes predictions about what the future occupancy states will be. Experiments performed using 2D and 3D laser data collected from crowded unstructured outdoor environments show that the proposed methodology can accurately predict occupancy states for areas of around 1000 m2at 10 Hz, making the proposed framework ideal for online applications under real-time constraints.

Mapping the occupancy of an environment is central for robot autonomy. Traditional occupancy grid maps discretise the environment into independent cells, neglecting important spatial correlations, and are unable to capture the continuous nature of the real world. With these drawbacks of grid maps in mind, Hilbert Maps (HM) and more recently Bayesian Hilbert Maps (BHMs), were introduced as a continuous representation of the environment. In this paper we propose a method to merge Bayesian Hilbert Maps built by a team of robots in a decentralised manner. The training of BHMs requires the inversion of a large covariance matrix, incurring cubic complexity. We introduce an approximation, Fast Bayesian Hilbert Maps (Fast-BHM), which reduces the time complexity to below quadratic. Such speed-ups allow the building and merging of Bayesian Hilbert Map models to be practical, opening the door for multi-robot Hilbert Map systems which can be much faster and more robust than an individual robot. By merging several individual Fast-BHMs in a decentralised manner we obtain a unified model of the environment which is itself a Fast-BHM. We conduct experiments to show that global Fast-BHM models do not deteriorate after repeated merging and training. We then empirically demonstrate, due to its the compact representation, fused Fast-BHMs outperform fusion methods involving discretising continuous representations, when the amount of information communicated is limited.

Mapping the occupancy level of an environment is important for a robot to navigate in unknown and unstructured environments. To this end, continuous occupancy mapping techniques which express the probability of a location as a function are used. In this work, we provide a theoretical analysis to compare and contrast the two major branches of Bayesian continuous occupancy mapping techniques---Gaussian process occupancy maps and Bayesian Hilbert maps---considering the fact that both utilize kernel functions to operate in a rich high-dimensional implicit feature space and use variational inference to learn parameters. Then, we extend the recent Bayesian Hilbert maps framework which is so far only used for stationary robots, to map large environments with moving robots. Finally, we propose convolution of kernels as a powerful tool to improve different aspects of continuous occupancy mapping. Our claims are also experimentally validated with both simulated and real-world datasets.

Robots often have to deal with the challenges of operating in dynamic and sometimes unpredictable environments. Although an occupancy map of the environment is sufficient for navigation of a mobile robot or manipulation tasks with a robotic arm in static environments, robots operating in dynamic environments demand richer information to improve robustness, efficiency, and safety. For instance, in path planning, it is important to know the direction of motion of dynamic objects at various locations of the environment for safer navigation or human-robot interaction. In this paper, we introduce directional statistics into robotic mapping to model circular data. Primarily, in collateral to occupancy grid maps, we propose directional grid maps to represent the location-wide long-term angular motion of the environment. Being highly representative, this defines a probability measure-field over the longitude-latitude space rather than a scalar-field or a vector-field. Withal, we further demonstrate how the same theory can be used to model angular variations in the spatial domain, temporal domain, and spatiotemporal domain. We carried out a series of experiments to validate the proposed models using a variety of robots having different sensors such as RGB cameras and LiDARs on simulated and real-world settings in both indoor and outdoor environments.

Understanding the dynamics of urban environments is crucial for path planning and safe navigation. However, the dynamics might be extremely complex making learning the environment an unfathomable task. Within the methods available for learning dynamic environments, dynamic Gaussian process occupancy maps (DGPOM) are very attractive because they can produce spatially-continuous occupancy maps taking into account neighborhood information, and provide probabilistic estimates, naturally inferring the uncertainty of predictions. Despite these properties, they are extremely slow, especially in dynamic mapping where the parameters of the map have to be updated as new data arrive from range sensors such as LiDARs. In this work, we leverage recent advancements in stochastic variational inference (SVI) to quickly learn dynamic areas in an online fashion. Further, we propose an information-driven technique to “intelligently” select inducing points required for SVI without relying on any object trackers which essentially improves computational time as well as robustness. These long-term occupancy maps entertain all attractive properties of DGPOM while the learning process is significantly faster, yet accurate. Our experiments with both simulation and real robot data on road intersections show a significant improvement in speed while maintaining a comparable or better accuracy.

Understanding and predicting how influenza propagates is vital to reduce its impact. In this paper we develop a nonparametric model based on Gaussian process (GP) regression to capture the complex spatial and temporal dependencies present in the data. A stochastic variational inference approach was adopted to address scalability. Rather than modeling the problem as a time-series as in many studies, we capture the space-time dependencies by combining different kernels. A kernel averaging technique which converts spatially-diffused point processes to an area process is proposed to model geographical distribution. Additionally, to accurately model the variable behavior of the time-series, the GP kernel is further modified to account for non-stationarity and seasonality. Experimental results on two datasets of state-wide US weekly flu-counts consisting of 19,698 and 89,474 data points, ranging over several years, illustrate the robustness of the model as a tool for further epidemiological investigations.

We consider the problem of building continuous occupancy representations in dynamic environments for robotics applications. The problem has hardly been discussed previously due to the complexity of patterns in urban environments, which have both spatial and temporal dependencies. We address the problem as learning a kernel classifier on an efficient feature space. The key novelty of our approach is the incorporation of variations in the time domain into the spatial domain. We propose a method to propagate motion uncertainty into the kernel using a hierarchical model. The main benefit of this approach is that it can directly predict the occupancy state of the map in the future from past observations, being a valuable tool for robot trajectory planning under uncertainty. Our approach preserves the main computational benefits of static Hilbert maps — using stochastic gradient descent for fast optimization of model parameters and incremental updates as new data are captured. Experiments conducted in road intersections of an urban environment demonstrated that spatio-temporal Hilbert maps can accurately model changes in the map while outperforming other techniques on various aspects.

Targeting and tracking in graphical user interfaces have been widely studied, but attempts to model targeted-tracking are few. Targeted-tracking is essentially a two component task of tracking followed by targeting, where either or both components may dominate depending on the levels of difficulty in each component. The applicability of an empirical model based on computer mouse use is unknown with respect to other devices. In order to confirm the model validity for other input devices, experiments were carried out using a mouse, a pen mouse, a touch screen, and a graphics tablet. Fourteen participants were tested on 48 experimental conditions that included four difficulty levels and 12 conditions with varying track width (P), track length (D), and target width (W). Movement time, error rate, index of performance, and throughput were compared. Repeated-measures ANOVA indicated that factors in the targeted-tracking model were significant (ρ2> 0.8) to the model, confirming the generality of the model. A principal component analysis showed that a mouse is relatively superior in terms of both movement time and error rate. Thus, the targeted-tracking model is an effective way to compare and evaluate input devices.