Anthony Corso

Lily Xu, Esther Rolf, Sara Beery et al. · mit

In this white paper, we synthesize key points made during presentations and discussions from the AI-Assisted Decision Making for Conservation workshop, hosted by the Center for Research on Computation and Society at Harvard University on October 20-21, 2022. We identify key open research questions in resource allocation, planning, and interventions for biodiversity conservation, highlighting conservation challenges that not only require AI solutions, but also require novel methodological advances. In addition to providing a summary of the workshop talks and discussions, we hope this document serves as a call-to-action to orient the expansion of algorithmic decision-making approaches to prioritize real-world conservation challenges, through collaborative efforts of ecologists, conservation decision-makers, and AI researchers.

Anthony Corso, Yizheng Wang, Markus Zechner et al.

Geological carbon capture and sequestration (CCS), where CO$_2$ is stored in subsurface formations, is a promising and scalable approach for reducing global emissions. However, if done incorrectly, it may lead to earthquakes and leakage of CO$_2$ back to the surface, harming both humans and the environment. These risks are exacerbated by the large amount of uncertainty in the structure of the storage formation. For these reasons, we propose that CCS operations be modeled as a partially observable Markov decision process (POMDP) and decisions be informed using automated planning algorithms. To this end, we develop a simplified model of CCS operations based on a 2D spillpoint analysis that retains many of the challenges and safety considerations of the real-world problem. We show how off-the-shelf POMDP solvers outperform expert baselines for safe CCS planning. This POMDP model can be used as a test bed to drive the development of novel decision-making algorithms for CCS operations.

Anthony Corso, David Karamadian, Romeo Valentin et al.

As machine learning (ML) systems increasingly permeate high-stakes settings such as healthcare, transportation, military, and national security, concerns regarding their reliability have emerged. Despite notable progress, the performance of these systems can significantly diminish due to adversarial attacks or environmental changes, leading to overconfident predictions, failures to detect input faults, and an inability to generalize in unexpected scenarios. This paper proposes a holistic assessment methodology for the reliability of ML systems. Our framework evaluates five key properties: in-distribution accuracy, distribution-shift robustness, adversarial robustness, calibration, and out-of-distribution detection. A reliability score is also introduced and used to assess the overall system reliability. To provide insights into the performance of different algorithmic approaches, we identify and categorize state-of-the-art techniques, then evaluate a selection on real-world tasks using our proposed reliability metrics and reliability score. Our analysis of over 500 models reveals that designing for one metric does not necessarily constrain others but certain algorithmic techniques can improve reliability across multiple metrics simultaneously. This study contributes to a more comprehensive understanding of ML reliability and provides a roadmap for future research and development.

Arec Jamgochian, Anthony Corso, Mykel J. Kochenderfer

Rather than augmenting rewards with penalties for undesired behavior, Constrained Partially Observable Markov Decision Processes (CPOMDPs) plan safely by imposing inviolable hard constraint value budgets. Previous work performing online planning for CPOMDPs has only been applied to discrete action and observation spaces. In this work, we propose algorithms for online CPOMDP planning for continuous state, action, and observation spaces by combining dual ascent with progressive widening. We empirically compare the effectiveness of our proposed algorithms on continuous CPOMDPs that model both toy and real-world safety-critical problems. Additionally, we compare against the use of online solvers for continuous unconstrained POMDPs that scalarize cost constraints into rewards, and investigate the effect of optimistic cost propagation.

Anthony Corso, Kyu-Young Kim, Shubh Gupta et al.

An important step in the design of autonomous systems is to evaluate the probability that a failure will occur. In safety-critical domains, the failure probability is extremely small so that the evaluation of a policy through Monte Carlo sampling is inefficient. Adaptive importance sampling approaches have been developed for rare event estimation but do not scale well to sequential systems with long horizons. In this work, we develop two adaptive importance sampling algorithms that can efficiently estimate the probability of rare events for sequential decision making systems. The basis for these algorithms is the minimization of the Kullback-Leibler divergence between a state-dependent proposal distribution and a target distribution over trajectories, but the resulting algorithms resemble policy gradient and value-based reinforcement learning. We apply multiple importance sampling to reduce the variance of our estimate and to address the issue of multi-modality in the optimal proposal distribution. We demonstrate our approach on a control task with both continuous and discrete actions spaces and show accuracy improvements over several baselines.

Arec Jamgochian, Hugo Buurmeijer, Kyle H. Wray et al.

Optimal plans in Constrained Partially Observable Markov Decision Processes (CPOMDPs) maximize reward objectives while satisfying hard cost constraints, generalizing safe planning under state and transition uncertainty. Unfortunately, online CPOMDP planning is extremely difficult in large or continuous problem domains. In many large robotic domains, hierarchical decomposition can simplify planning by using tools for low-level control given high-level action primitives (options). We introduce Constrained Options Belief Tree Search (COBeTS) to leverage this hierarchy and scale online search-based CPOMDP planning to large robotic problems. We show that if primitive option controllers are defined to satisfy assigned constraint budgets, then COBeTS will satisfy constraints anytime. Otherwise, COBeTS will guide the search towards a safe sequence of option primitives, and hierarchical monitoring can be used to achieve runtime safety. We demonstrate COBeTS in several safety-critical, constrained partially observable robotic domains, showing that it can plan successfully in continuous CPOMDPs while non-hierarchical baselines cannot.

Harrison Delecki, Anthony Corso, Mykel J. Kochenderfer

Estimating the distribution over failures is a key step in validating autonomous systems. Existing approaches focus on finding failures for a small range of initial conditions or make restrictive assumptions about the properties of the system under test. We frame estimating the distribution over failure trajectories for sequential systems as Bayesian inference. Our model-based approach represents the distribution over failure trajectories using rollouts of system dynamics and computes trajectory gradients using automatic differentiation. Our approach is demonstrated in an inverted pendulum control system, an autonomous vehicle driving scenario, and a partially observable lunar lander. Sampling is performed using an off-the-shelf implementation of Hamiltonian Monte Carlo with multiple chains to capture multimodality and gradient smoothing for safe trajectories. In all experiments, we observed improvements in sample efficiency and parameter space coverage compared to black-box baseline approaches. This work is open sourced.

Max Lamparth, Anthony Corso, Jacob Ganz et al. · stanford

To some, the advent of artificial intelligence (AI) promises better decision-making and increased military effectiveness while reducing the influence of human error and emotions. However, there is still debate about how AI systems, especially large language models (LLMs) that can be applied to many tasks, behave compared to humans in high-stakes military decision-making scenarios with the potential for increased risks towards escalation. To test this potential and scrutinize the use of LLMs for such purposes, we use a new wargame experiment with 214 national security experts designed to examine crisis escalation in a fictional U.S.-China scenario and compare the behavior of human player teams to LLM-simulated team responses in separate simulations. Here, we find that the LLM-simulated responses can be more aggressive and significantly affected by changes in the scenario. We show a considerable high-level agreement in the LLM and human responses and significant quantitative and qualitative differences in individual actions and strategic tendencies. These differences depend on intrinsic biases in LLMs regarding the appropriate level of violence following strategic instructions, the choice of LLM, and whether the LLMs are tasked to decide for a team of players directly or first to simulate dialog between a team of players. When simulating the dialog, the discussions lack quality and maintain a farcical harmony. The LLM simulations cannot account for human player characteristics, showing no significant difference even for extreme traits, such as "pacifist" or "aggressive sociopath." When probing behavioral consistency across individual moves of the simulation, the tested LLMs deviated from each other but generally showed somewhat consistent behavior. Our results motivate policymakers to be cautious before granting autonomy or following AI-based strategy recommendations.

Robert J. Moss, Arec Jamgochian, Johannes Fischer et al.

To plan safely in uncertain environments, agents must balance utility with safety constraints. Safe planning problems can be modeled as a chance-constrained partially observable Markov decision process (CC-POMDP) and solutions often use expensive rollouts or heuristics to estimate the optimal value and action-selection policy. This work introduces the ConstrainedZero policy iteration algorithm that solves CC-POMDPs in belief space by learning neural network approximations of the optimal value and policy with an additional network head that estimates the failure probability given a belief. This failure probability guides safe action selection during online Monte Carlo tree search (MCTS). To avoid overemphasizing search based on the failure estimates, we introduce $Δ$-MCTS, which uses adaptive conformal inference to update the failure threshold during planning. The approach is tested on a safety-critical POMDP benchmark, an aircraft collision avoidance system, and the sustainability problem of safe CO$_2$ storage. Results show that by separating safety constraints from the objective we can achieve a target level of safety without optimizing the balance between rewards and costs.

Nick DiSanto, Anthony Corso, Benjamin Sanders et al.

While transformers have pioneered attention-driven architectures as a cornerstone of language modeling, their dependence on explicitly contextual information underscores limitations in their abilities to tacitly learn overarching textual themes. This study challenges the heuristic paradigm of performance benchmarking by investigating social media data as a source of distributed patterns. In stark contrast to networks that rely on capturing complex long-term dependencies, models of online data inherently lack structure and are forced to detect latent structures in the aggregate. To properly represent these abstract relationships, this research dissects empirical social media corpora into their elemental components, analyzing over two billion tweets across population-dense locations. We create Bag-of-Word embedding specific to each city and compare their respective representations. This finds that even amidst noisy data, geographic location has a considerable influence on online communication, and that hidden insights can be uncovered without the crutch of advanced algorithms. This evidence presents valuable geospatial implications in social science and challenges the notion that intricate models are prerequisites for pattern recognition in natural language. This aligns with the evolving landscape that questions the embrace of absolute interpretability over abstract understanding and bridges the divide between sophisticated frameworks and intangible relationships.

Mansur Arief, Yasmine Alonso, CJ Oshiro et al.

The world is entering an unprecedented period of critical mineral demand, driven by the global transition to renewable energy technologies and electric vehicles. This transition presents unique challenges in mineral resource development, particularly due to geological uncertainty-a key characteristic that traditional supply chain optimization approaches do not adequately address. To tackle this challenge, we propose a novel application of Partially Observable Markov Decision Processes (POMDPs) that optimizes critical mineral sourcing decisions while explicitly accounting for the dynamic nature of geological uncertainty. Through a case study of the U.S. lithium supply chain, we demonstrate that POMDP-based policies achieve superior outcomes compared to traditional approaches, especially when initial reserve estimates are imperfect. Our framework provides quantitative insights for balancing domestic resource development with international supply diversification, offering policymakers a systematic approach to strategic decision-making in critical mineral supply chains.

John Mern, Anthony Corso, Damian Burch et al.

Optimal Bayesian decision making on what geoscientific data to acquire requires stating a prior model of uncertainty. Data acquisition is then optimized by reducing uncertainty on some property of interest maximally, and on average. In the context of exploration, very few, sometimes no data at all, is available prior to data acquisition planning. The prior model therefore needs to include human interpretations on the nature of spatial variability, or on analogue data deemed relevant for the area being explored. In mineral exploration, for example, humans may rely on conceptual models on the genesis of the mineralization to define multiple hypotheses, each representing a specific spatial variability of mineralization. More often than not, after the data is acquired, all of the stated hypotheses may be proven incorrect, i.e. falsified, hence prior hypotheses need to be revised, or additional hypotheses generated. Planning data acquisition under wrong geological priors is likely to be inefficient since the estimated uncertainty on the target property is incorrect, hence uncertainty may not be reduced at all. In this paper, we develop an intelligent agent based on partially observable Markov decision processes that plans optimally in the case of multiple geological or geoscientific hypotheses on the nature of spatial variability. Additionally, the artificial intelligence is equipped with a method that allows detecting, early on, whether the human stated hypotheses are incorrect, thereby saving considerable expense in data acquisition. Our approach is tested on a sediment-hosted copper deposit, and the algorithm presented has aided in the characterization of an ultra high-grade deposit in Zambia in 2023.

Harrison Delecki, Marc R. Schlichting, Mansur Arief et al.

Validating safety-critical autonomous systems in high-dimensional domains such as robotics presents a significant challenge. Existing black-box approaches based on Markov chain Monte Carlo may require an enormous number of samples, while methods based on importance sampling often rely on simple parametric families that may struggle to represent the distribution over failures. We propose to sample the distribution over failures using a conditional denoising diffusion model, which has shown success in complex high-dimensional problems such as robotic task planning. We iteratively train a diffusion model to produce state trajectories closer to failure. We demonstrate the effectiveness of our approach on high-dimensional robotic validation tasks, improving sample efficiency and mode coverage compared to existing black-box techniques.

Robert J. Moss, Anthony Corso, Jef Caers et al.

Real-world planning problems, including autonomous driving and sustainable energy applications like carbon storage and resource exploration, have recently been modeled as partially observable Markov decision processes (POMDPs) and solved using approximate methods. To solve high-dimensional POMDPs in practice, state-of-the-art methods use online planning with problem-specific heuristics to reduce planning horizons and make the problems tractable. Algorithms that learn approximations to replace heuristics have recently found success in large-scale fully observable domains. The key insight is the combination of online Monte Carlo tree search with offline neural network approximations of the optimal policy and value function. In this work, we bring this insight to partially observable domains and propose BetaZero, a belief-state planning algorithm for high-dimensional POMDPs. BetaZero learns offline approximations that replace heuristics to enable online decision making in long-horizon problems. We address several challenges inherent in large-scale partially observable domains; namely challenges of transitioning in stochastic environments, prioritizing action branching with a limited search budget, and representing beliefs as input to the network. To formalize the use of all limited search information, we train against a novel $Q$-weighted visit counts policy. We test BetaZero on various well-established POMDP benchmarks found in the literature and a real-world problem of critical mineral exploration. Experiments show that BetaZero outperforms state-of-the-art POMDP solvers on a variety of tasks.

Chelsea Sidrane, Sydney Katz, Anthony Corso et al.

Inverse problems exist in a wide variety of physical domains from aerospace engineering to medical imaging. The goal is to infer the underlying state from a set of observations. When the forward model that produced the observations is nonlinear and stochastic, solving the inverse problem is very challenging. Neural networks are an appealing solution for solving inverse problems as they can be trained from noisy data and once trained are computationally efficient to run. However, inverse model neural networks do not have guarantees of correctness built-in, which makes them unreliable for use in safety and accuracy-critical contexts. In this work we introduce a method for verifying the correctness of inverse model neural networks. Our approach is to overapproximate a nonlinear, stochastic forward model with piecewise linear constraints and encode both the overapproximate forward model and the neural network inverse model as a mixed-integer program. We demonstrate this verification procedure on a real-world airplane fuel gauge case study. The ability to verify and consequently trust inverse model neural networks allows their use in a wide variety of contexts, from aerospace to medicine.

Anthony Corso, Mykel J. Kochenderfer

Safety validation is important during the development of safety-critical autonomous systems but can require significant computational effort. Existing algorithms often start from scratch each time the system under test changes. We apply transfer learning to improve the efficiency of reinforcement learning based safety validation algorithms when applied to related systems. Knowledge from previous safety validation tasks is encoded through the action value function and transferred to future tasks with a learned set of attention weights. Including a learned state and action value transformation for each source task can improve performance even when systems have substantially different failure modes. We conduct experiments on safety validation tasks in gridworld and autonomous driving scenarios. We show that transfer learning can improve the initial and final performance of validation algorithms and reduce the number of training steps.

Anthony Corso, Robert J. Moss, Mark Koren et al.

Autonomous cyber-physical systems (CPS) can improve safety and efficiency for safety-critical applications, but require rigorous testing before deployment. The complexity of these systems often precludes the use of formal verification and real-world testing can be too dangerous during development. Therefore, simulation-based techniques have been developed that treat the system under test as a black box operating in a simulated environment. Safety validation tasks include finding disturbances in the environment that cause the system to fail (falsification), finding the most-likely failure, and estimating the probability that the system fails. Motivated by the prevalence of safety-critical artificial intelligence, this work provides a survey of state-of-the-art safety validation techniques for CPS with a focus on applied algorithms and their modifications for the safety validation problem. We present and discuss algorithms in the domains of optimization, path planning, reinforcement learning, and importance sampling. Problem decomposition techniques are presented to help scale algorithms to large state spaces, which are common for CPS. A brief overview of safety-critical applications is given, including autonomous vehicles and aircraft collision avoidance systems. Finally, we present a survey of existing academic and commercially available safety validation tools.

Anthony Corso, Mykel J. Kochenderfer

An open problem for autonomous driving is how to validate the safety of an autonomous vehicle in simulation. Automated testing procedures can find failures of an autonomous system but these failures may be difficult to interpret due to their high dimensionality and may be so unlikely as to not be important. This work describes an approach for finding interpretable failures of an autonomous system. The failures are described by signal temporal logic expressions that can be understood by a human, and are optimized to produce failures that have high likelihood. Our methodology is demonstrated for the safety validation of an autonomous vehicle in the context of an unprotected left turn and a crosswalk with a pedestrian. Compared to a baseline importance sampling approach, our methodology finds more failures with higher likelihood while retaining interpretability.

Anthony Corso, Ritchie Lee, Mykel J. Kochenderfer

An open question in autonomous driving is how best to use simulation to validate the safety of autonomous vehicles. Existing techniques rely on simulated rollouts, which can be inefficient for finding rare failure events, while other techniques are designed to only discover a single failure. In this work, we present a new safety validation approach that attempts to estimate the distribution over failures of an autonomous policy using approximate dynamic programming. Knowledge of this distribution allows for the efficient discovery of many failure examples. To address the problem of scalability, we decompose complex driving scenarios into subproblems consisting of only the ego vehicle and one other vehicle. These subproblems can be solved with approximate dynamic programming and their solutions are recombined to approximate the solution to the full scenario. We apply our approach to a simple two-vehicle scenario to demonstrate the technique as well as a more complex five-vehicle scenario to demonstrate scalability. In both experiments, we observed an increase in the number of failures discovered compared to baseline approaches.

Mark Koren, Anthony Corso, Mykel J. Kochenderfer

Validation is a key challenge in the search for safe autonomy. Simulations are often either too simple to provide robust validation, or too complex to tractably compute. Therefore, approximate validation methods are needed to tractably find failures without unsafe simplifications. This paper presents the theory behind one such black-box approach: adaptive stress testing (AST). We also provide three examples of validation problems formulated to work with AST.

Anthony Corso, Peter Du, Katherine Driggs-Campbell et al.

Determining possible failure scenarios is a critical step in the evaluation of autonomous vehicle systems. Real-world vehicle testing is commonly employed for autonomous vehicle validation, but the costs and time requirements are high. Consequently, simulation-driven methods such as Adaptive Stress Testing (AST) have been proposed to aid in validation. AST formulates the problem of finding the most likely failure scenarios as a Markov decision process, which can be solved using reinforcement learning. In practice, AST tends to find scenarios where failure is unavoidable and tends to repeatedly discover the same types of failures of a system. This work addresses these issues by encoding domain relevant information into the search procedure. With this modification, the AST method discovers a larger and more expressive subset of the failure space when compared to the original AST formulation. We show that our approach is able to identify useful failure scenarios of an autonomous vehicle policy.