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 and unnecessary conflicts. To test this potential and scrutinize the use of LLMs for such purposes, we use a new wargame experiment with 107 national security experts designed to examine crisis escalation in a fictional US-China scenario and compare the behavior of human player teams to LLM-simulated team responses in separate simulations. Wargames have a long history in the development of military strategy and the response of nations to threats or attacks. 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.

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.

One of the bottlenecks of training autonomous vehicle (AV) agents is the variability of training environments. Since learning optimal policies for unseen environments is often very costly and requires substantial data collection, it becomes computationally intractable to train the agent on every possible environment or task the AV may encounter.This paper introduces a zero-shot filtering approach to interpolate learned policies of past experiences to generalize to unseen ones. We use an experience kernel to correlate environments. These correlations are then exploited to produce policies for new tasks or environments from learned policies. We demonstrate our methods on an autonomous vehicle driving through T-intersections with different characteristics, where its behavior is modeled as a partially observable Markov decision process (POMDP). We first construct compact representations of learned policies for POMDPs with unknown transition functions given a dataset of sequential actions and observations. Then, we filter parameterized policies of previously visited environments to generate policies to new, unseen environments. We demonstrate our approaches on both an actual AV and a high-fidelity simulator. Results indicate that our experience filter offers a fast, low-effort, and near-optimal solution to create policies for tasks or environments never seen before. Furthermore, the generated new policies outperform the policy learned using the entire data collected from past environments, suggesting that the correlation among different environments can be exploited and irrelevant ones can be filtered out.

Designing robust machine learning systems remains an open problem, and there is a need for benchmark problems that cover both environmental changes and evaluation on a downstream task. In this work, we introduce AVOIDDS, a realistic object detection benchmark for the vision-based aircraft detect-and-avoid problem. We provide a labeled dataset consisting of 72,000 photorealistic images of intruder aircraft with various lighting conditions, weather conditions, relative geometries, and geographic locations. We also provide an interface that evaluates trained models on slices of this dataset to identify changes in performance with respect to changing environmental conditions. Finally, we implement a fully-integrated, closed-loop simulator of the vision-based detect-and-avoid problem to evaluate trained models with respect to the downstream collision avoidance task. This benchmark will enable further research in the design of robust machine learning systems for use in safety-critical applications. The AVOIDDS dataset and code are publicly available at https://purl.stanford.edu/hj293cv5980 and https://github.com/sisl/VisionBasedAircraftDAA, respectively.

Modern autonomous systems rely on perception modules to process complex sensor measurements into state estimates. These estimates are then passed to a controller, which uses them to make safety-critical decisions. It is therefore important that we design perception systems to minimize errors that reduce the overall safety of the system. We develop a risk-driven approach to designing perception systems that accounts for the effect of perceptual errors on the performance of the fully-integrated, closed-loop system. We formulate a risk function to quantify the effect of a given perceptual error on overall safety, and show how we can use it to design safer perception systems by including a risk-dependent term in the loss function and generating training data in risk-sensitive regions. We evaluate our techniques on a realistic vision-based aircraft detect and avoid application and show that risk-driven design reduces collision risk by 37% over a baseline system.

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.

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.