From Foresight to Forethought: VLM-In-the-Loop Policy Steering via Latent Alignment

While generative robot policies have demonstrated significant potential in learning complex, multimodal behaviors from demonstrations, they still exhibit diverse failures at deployment-time. Policy steering offers an elegant solution to reducing the chance of failure by using an external verifier to select from low-level actions proposed by an imperfect generative policy. Here, one might hope to use a Vision Language Model (VLM) as a verifier, leveraging its open-world reasoning capabilities. However, off-the-shelf VLMs struggle to understand the consequences of low-level robot actions as they are represented fundamentally differently than the text and images the VLM was trained on. In response, we propose FOREWARN, a novel framework to unlock the potential of VLMs as open-vocabulary verifiers for runtime policy steering. Our key idea is to decouple the VLM's burden of predicting action outcomes (foresight) from evaluation (forethought). For foresight, we leverage a latent world model to imagine future latent states given diverse low-level action plans. For forethought, we align the VLM with these predicted latent states to reason about the consequences of actions in its native representation--natural language--and effectively filter proposed plans. We validate our framework across diverse robotic manipulation tasks, demonstrating its ability to bridge representational gaps and provide robust, generalizable policy steering.

Generalizing Safety Beyond Collision-Avoidance via Latent-Space Reachability Analysis

Hamilton-Jacobi (HJ) reachability is a rigorous mathematical framework that enables robots to simultaneously detect unsafe states and generate actions that prevent future failures. While in theory, HJ reachability can synthesize safe controllers for nonlinear systems and nonconvex constraints, in practice, it has been limited to hand-engineered collision-avoidance constraints modeled via low-dimensional state-space representations and first-principles dynamics. In this work, our goal is to generalize safe robot controllers to prevent failures that are hard -- if not impossible -- to write down by hand, but can be intuitively identified from high-dimensional observations: for example, spilling the contents of a bag. We propose Latent Safety Filters, a latent-space generalization of HJ reachability that tractably operates directly on raw observation data (e.g., RGB images) by performing safety analysis in the latent embedding space of a generative world model. This transforms nuanced constraint specification to a classification problem in latent space and enables reasoning about dynamical consequences that are hard to simulate. In simulation and hardware experiments, we use Latent Safety Filters to safeguard arbitrary policies (from generative policies to direct teleoperation) from complex safety hazards, like preventing a Franka Research 3 manipulator from spilling the contents of a bag or toppling cluttered objects.

Your Learned Constraint is Secretly a Backward Reachable Tube

Inverse Constraint Learning (ICL) is the problem of inferring constraints from safe (i.e., constraint-satisfying) demonstrations. The hope is that these inferred constraints can then be used downstream to search for safe policies for new tasks and, potentially, under different dynamics. Our paper explores the question of what mathematical entity ICL recovers. Somewhat surprisingly, we show that both in theory and in practice, ICL recovers the set of states where failure is inevitable, rather than the set of states where failure has already happened. In the language of safe control, this means we recover a backwards reachable tube (BRT) rather than a failure set. In contrast to the failure set, the BRT depends on the dynamics of the data collection system. We discuss the implications of the dynamics-conditionedness of the recovered constraint on both the sample-efficiency of policy search and the transferability of learned constraints.

Maximizing Alignment with Minimal Feedback: Efficiently Learning Rewards for Visuomotor Robot Policy Alignment

Visuomotor robot policies, increasingly pre-trained on large-scale datasets, promise significant advancements across robotics domains. However, aligning these policies with end-user preferences remains a challenge, particularly when the preferences are hard to specify. While reinforcement learning from human feedback (RLHF) has become the predominant mechanism for alignment in non-embodied domains like large language models, it has not seen the same success in aligning visuomotor policies due to the prohibitive amount of human feedback required to learn visual reward functions. To address this limitation, we propose Representation-Aligned Preference-based Learning (RAPL), an observation-only method for learning visual rewards from significantly less human preference feedback. Unlike traditional RLHF, RAPL focuses human feedback on fine-tuning pre-trained vision encoders to align with the end-user's visual representation and then constructs a dense visual reward via feature matching in this aligned representation space. We first validate RAPL through simulation experiments in the X-Magical benchmark and Franka Panda robotic manipulation, demonstrating that it can learn rewards aligned with human preferences, more efficiently uses preference data, and generalizes across robot embodiments. Finally, our hardware experiments align pre-trained Diffusion Policies for three object manipulation tasks. We find that RAPL can fine-tune these policies with 5x less real human preference data, taking the first step towards minimizing human feedback while maximizing visuomotor robot policy alignment.

Agent-to-Sim: Learning Interactive Behavior Models from Casual Longitudinal Videos

We present Agent-to-Sim (ATS), a framework for learning interactive behavior models of 3D agents from casual longitudinal video collections. Different from prior works that rely on marker-based tracking and multiview cameras, ATS learns natural behaviors of animal and human agents non-invasively through video observations recorded over a long time-span (e.g., a month) in a single environment. Modeling 3D behavior of an agent requires persistent 3D tracking (e.g., knowing which point corresponds to which) over a long time period. To obtain such data, we develop a coarse-to-fine registration method that tracks the agent and the camera over time through a canonical 3D space, resulting in a complete and persistent spacetime 4D representation. We then train a generative model of agent behaviors using paired data of perception and motion of an agent queried from the 4D reconstruction. ATS enables real-to-sim transfer from video recordings of an agent to an interactive behavior simulator. We demonstrate results on pets (e.g., cat, dog, bunny) and human given monocular RGBD videos captured by a smartphone.

Conformalized Interactive Imitation Learning: Handling Expert Shift and Intermittent Feedback

In interactive imitation learning (IL), uncertainty quantification offers a way for the learner (i.e. robot) to contend with distribution shifts encountered during deployment by actively seeking additional feedback from an expert (i.e. human) online. Prior works use mechanisms like ensemble disagreement or Monte Carlo dropout to quantify when black-box IL policies are uncertain; however, these approaches can lead to overconfident estimates when faced with deployment-time distribution shifts. Instead, we contend that we need uncertainty quantification algorithms that can leverage the expert human feedback received during deployment time to adapt the robot's uncertainty online. To tackle this, we draw upon online conformal prediction, a distribution-free method for constructing prediction intervals online given a stream of ground-truth labels. Human labels, however, are intermittent in the interactive IL setting. Thus, from the conformal prediction side, we introduce a novel uncertainty quantification algorithm called intermittent quantile tracking (IQT) that leverages a probabilistic model of intermittent labels, maintains asymptotic coverage guarantees, and empirically achieves desired coverage levels. From the interactive IL side, we develop ConformalDAgger, a new approach wherein the robot uses prediction intervals calibrated by IQT as a reliable measure of deployment-time uncertainty to actively query for more expert feedback. We compare ConformalDAgger to prior uncertainty-aware DAgger methods in scenarios where the distribution shift is (and isn't) present because of changes in the expert's policy. We find that in simulated and hardware deployments on a 7DOF robotic manipulator, ConformalDAgger detects high uncertainty when the expert shifts and increases the number of interventions compared to baselines, allowing the robot to more quickly learn the new behavior.

Updating Robot Safety Representations Online from Natural Language Feedback

Robots must operate safely when deployed in novel and human-centered environments, like homes. Current safe control approaches typically assume that the safety constraints are known a priori, and thus, the robot can pre-compute a corresponding safety controller. While this may make sense for some safety constraints (e.g., avoiding collision with walls by analyzing a floor plan), other constraints are more complex (e.g., spills), inherently personal, context-dependent, and can only be identified at deployment time when the robot is interacting in a specific environment and with a specific person (e.g., fragile objects, expensive rugs). Here, language provides a flexible mechanism to communicate these evolving safety constraints to the robot. In this work, we use vision language models (VLMs) to interpret language feedback and the robot's image observations to continuously update the robot's representation of safety constraints. With these inferred constraints, we update a Hamilton-Jacobi reachability safety controller online via efficient warm-starting techniques. Through simulation and hardware experiments, we demonstrate the robot's ability to infer and respect language-based safety constraints with the proposed approach.

Robots that Learn to Safely Influence via Prediction-Informed Reach-Avoid Dynamic Games

Robots can influence people to accomplish their tasks more efficiently: autonomous cars can inch forward at an intersection to pass through, and tabletop manipulators can go for an object on the table first. However, a robot's ability to influence can also compromise the safety of nearby people if naively executed. In this work, we pose and solve a novel robust reach-avoid dynamic game which enables robots to be maximally influential, but only when a safety backup control exists. On the human side, we model the human's behavior as goal-driven but conditioned on the robot's plan, enabling us to capture influence. On the robot side, we solve the dynamic game in the joint physical and belief space, enabling the robot to reason about how its uncertainty in human behavior will evolve over time. We instantiate our method, called SLIDE (Safely Leveraging Influence in Dynamic Environments), in a high-dimensional (39-D) simulated human-robot collaborative manipulation task solved via offline game-theoretic reinforcement learning. We compare our approach to a robust baseline that treats the human as a worst-case adversary, a safety controller that does not explicitly reason about influence, and an energy-function-based safety shield. We find that SLIDE consistently enables the robot to leverage the influence it has on the human when it is safe to do so, ultimately allowing the robot to be less conservative while still ensuring a high safety rate during task execution.

Robots that Suggest Safe Alternatives

Goal-conditioned policies, such as those learned via imitation learning, provide an easy way for humans to influence what tasks robots accomplish. However, these robot policies are not guaranteed to execute safely or to succeed when faced with out-of-distribution requests. In this work, we enable robots to know when they can confidently execute a user's desired goal, and automatically suggest safe alternatives when they cannot. Our approach is inspired by control-theoretic safety filtering, wherein a safety filter minimally adjusts a robot's candidate action to be safe. Our key idea is to pose alternative suggestion as a safe control problem in goal space, rather than in action space. Offline, we use reachability analysis to compute a goal-parameterized reach-avoid value network which quantifies the safety and liveness of the robot's pre-trained policy. Online, our robot uses the reach-avoid value network as a safety filter, monitoring the human's given goal and actively suggesting alternatives that are similar but meet the safety specification. We demonstrate our Safe ALTernatives (SALT) framework in simulation experiments with indoor navigation and Franka Panda tabletop manipulation, and with both discrete and continuous goal representations. We find that SALT is able to learn to predict successful and failed closed-loop executions, is a less pessimistic monitor than open-loop uncertainty quantification, and proposes alternatives that consistently align with those people find acceptable.

Conformalized Teleoperation: Confidently Mapping Human Inputs to High-Dimensional Robot Actions

Assistive robotic arms often have more degrees-of-freedom than a human teleoperator can control with a low-dimensional input, like a joystick. To overcome this challenge, existing approaches use data-driven methods to learn a mapping from low-dimensional human inputs to high-dimensional robot actions. However, determining if such a black-box mapping can confidently infer a user's intended high-dimensional action from low-dimensional inputs remains an open problem. Our key idea is to adapt the assistive map at training time to additionally estimate high-dimensional action quantiles, and then calibrate these quantiles via rigorous uncertainty quantification methods. Specifically, we leverage adaptive conformal prediction which adjusts the intervals over time, reducing the uncertainty bounds when the mapping is performant and increasing the bounds when the mapping consistently mis-predicts. Furthermore, we propose an uncertainty-interval-based mechanism for detecting high-uncertainty user inputs and robot states. We evaluate the efficacy of our proposed approach in a 2D assistive navigation task and two 7DOF Kinova Jaco tasks involving assistive cup grasping and goal reaching. Our findings demonstrate that conformalized assistive teleoperation manages to detect (but not differentiate between) high uncertainty induced by diverse preferences and induced by low-precision trajectories in the mapping's training dataset. On the whole, we see this work as a key step towards enabling robots to quantify their own uncertainty and proactively seek intervention when needed.