Search for: All records

Recent works have shown that foundational safe control methods, such as Hamilton-Jacobi (HJ) reachability analysis, can be applied in the latent space of world models. While this enables the synthesis of latent safety filters for hard-to-model vision-based tasks, they assume that the safety constraint is known a priori and remains fixed during deployment, limiting the safety filter's adaptability across scenarios. To address this, we propose constraint-parameterized latent safety filters that can adapt to user-specified safety constraints at runtime. Our key idea is to define safety constraints by conditioning on an encoding of an image that represents a constraint, using a latent-space similarity measure. The notion of similarity to failure is aligned in a principled way through conformal calibration, which controls how closely the system may approach the constraint representation. The parameterized safety filter is trained entirely within the world model's imagination, treating any image seen by the model as a potential test-time constraint, thereby enabling runtime adaptation to arbitrary safety constraints. In simulation and hardware experiments on vision-based control tasks with a Franka manipulator, we show that our method adapts at runtime by conditioning on the encoding of user-specified constraint images, without sacrificing performance. 

Recent advances in generative world models have enabled classical safe control methods, such as Hamilton-Jacobi (HJ) reachability, to generalize to complex robotic systems operating directly from high-dimensional sensor observations. However, obtaining comprehensive coverage of all safety-critical scenarios during world model training is extremely challenging. As a result, latent safety filters built on top of these models may miss novel hazards and even fail to prevent known ones, overconfidently misclassifying risky out-of-distribution (OOD) situations as safe. To address this, we introduce an uncertainty-aware latent safety filter that proactively steers robots away from both known and unseen failures. Our key idea is to use the world model's epistemic uncertainty as a proxy for identifying unseen potential hazards. We propose a principled method to detect OOD world model predictions by calibrating an uncertainty threshold via conformal prediction. By performing reachability analysis in an augmented state space-spanning both the latent representation and the epistemic uncertainty-we synthesize a latent safety filter that can reliably safeguard arbitrary policies from both known and unseen safety hazards. In simulation and hardware experiments on vision-based control tasks with a Franka manipulator, we show that our uncertainty-aware safety filter preemptively detects potential unsafe scenarios and reliably proposes safe, in-distribution actions. 

The oral route is the most common choice for drug administration because of several advantages, such as convenience, low cost, and high patient compliance, and the demand and investment in research and development for oral drugs continue to grow. The rate of dissolution and gastric emptying of the dissolved active pharmaceutical ingredient (API) into the duodenum is modulated by gastric motility, physical properties of the pill, and the contents of the stomach, but current in vitro procedures for assessing dissolution of oral drugs are limited in their ability to recapitulate this process. This is particularly relevant for disease conditions, such as gastroparesis, that alter the anatomy and/or physiology of the stomach. In silico models of gastric biomechanics offer the potential for overcoming these limitations of existing methods. In the current study, we employ a biomimetic in silico simulator based on the realistic anatomy and morphology of the stomach (referred to as “StomachSim”) to investigate and quantify the effect of body posture and stomach motility on drug bioavailability. The simulations show that changes in posture can potentially have a significant (up to 83%) effect on the emptying rate of the API into the duodenum. Similarly, a reduction in antral contractility associated with gastroparesis can also be found to significantly reduce the dissolution of the pill as well as emptying of the API into the duodenum. The simulations show that for an equivalent motility index, the reduction in gastric emptying due to neuropathic gastroparesis is larger by a factor of about five compared to myopathic gastroparesis. 

Productivity tracking tools often determine productivity based on the time interacting with work-related applications. To deconstruct productivity’s diverse and nebulous nature, we investigate how knowledge workers conceptualize personal productivity and delimit productive tasks in both work and non-work contexts. We report a 2-week diary study followed by a semi-structured interview with 24 knowledge workers. Participants captured productive activities and provided the rationale for why the activities were assessed to be productive. They reported a wide range of productive activities beyond typical desk-bound work—ranging from having a personal conversation with dad to getting a haircut. We found six themes that characterize the productivity assessment—work product, time management, worker’s state, attitude toward work, impact & benefit, and compound task—and identified how participants interleaved multiple facets when assessing their productivity. We discuss how these findings could inform the design of a comprehensive productivity tracking system that covers a wide range of productive activities. 

The binary black hole signal GW250114, the loudest gravitational wave detected to date, offers a unique opportunity to test Einstein’s general relativity (GR) in the high-velocity, strong-gravity regime and probe whether the remnant conforms to the Kerr metric. Upon perturbation, black holes emit a spectrum of damped sinusoids with specific, complex frequencies. Our analysis of the postmerger signal shows that at least two quasinormal modes are required to explain the data, with the most damped remaining statistically significant for about one cycle. We probe the remnant’s Kerr nature by constraining the spectroscopic pattern of the dominant quadrupolar ( $\ell =m=2$) mode and its first overtone to match the Kerr prediction to tens of percent at multiple postpeak times. The measured mode amplitudes and phases agree with a numerical-relativity simulation having parameters close to GW250114. By fitting a parametrized waveform that incorporates the full inspiral-merger-ringdown sequence, we constrain the fundamental $(\ell =m=4)$mode to tens of percent and bound the quadrupolar frequency to within a few percent of the GR prediction. We perform a suite of tests—spanning inspiral, merger, and ringdown—finding constraints that are comparable to, and in some cases 2–3 times more stringent than those obtained by combining dozens of events in the fourth Gravitational-Wave Transient Catalog. These results constitute the most stringent single-event verification of GR and the Kerr nature of black holes to date, and outline the power of black-hole spectroscopy for future gravitational-wave observations.