Fanjun Bu

Jan Ondras, Abrar Anwar, Tong Wu et al.

We develop data-driven models to predict when a robot should feed during social dining scenarios. Being able to eat independently with friends and family is considered one of the most memorable and important activities for people with mobility limitations. While existing robotic systems for feeding people with mobility limitations focus on solitary dining, commensality, the act of eating together, is often the practice of choice. Sharing meals with others introduces the problem of socially appropriate bite timing for a robot, i.e. the appropriate timing for the robot to feed without disrupting the social dynamics of a shared meal. Our key insight is that bite timing strategies that take into account the delicate balance of social cues can lead to seamless interactions during robot-assisted feeding in a social dining scenario. We approach this problem by collecting a Human-Human Commensality Dataset (HHCD) containing 30 groups of three people eating together. We use this dataset to analyze human-human commensality behaviors and develop bite timing prediction models in social dining scenarios. We also transfer these models to human-robot commensality scenarios. Our user studies show that prediction improves when our algorithm uses multimodal social signaling cues between diners to model bite timing. The HHCD dataset, videos of user studies, and code are available at https://emprise.cs.cornell.edu/hrcom/

Fanjun Bu, Chenyang Yuan, Hiroshi Yasuda

Generative world models offer a compelling foundation for augmented-reality (AR) applications: by predicting future image sequences that incorporate deliberate visual edits, they enable temporally coherent, augmented future frames that can be computed ahead of time and cached, avoiding per-frame rendering from scratch in real time. In this work, we present SEGAR, a preliminary framework that combines a diffusion-based world model with a selective correction stage to support this vision. The world model generates augmented future frames with region-specific edits while preserving others, and the correction stage subsequently aligns safety-critical regions with real-world observations while preserving intended augmentations elsewhere. We demonstrate this pipeline in driving scenarios as a representative setting where semantic region structure is well defined and real-world feedback is readily available. We view this as an early step toward generative world models as practical AR infrastructure, where future frames can be generated, cached, and selectively corrected on demand.

Shima Rahimi Moghaddam, Fanjun Bu, Christopher J. Honey

In the human brain, internal states are often correlated over time (due to local recurrence and other intrinsic circuit properties), punctuated by abrupt transitions. At first glance, temporal smoothness of internal states presents a problem for learning input-output mappings (e.g. category labels for images), because the internal representation of the input will contain a mixture of current input and prior inputs. However, when training with naturalistic data (e.g. movies) there is also temporal autocorrelation in the input. How does the temporal "smoothness" of internal states affect the efficiency of learning when the training data are also temporally smooth? How does it affect the kinds of representations that are learned? We found that, when trained with temporally smooth data, "slow" neural networks (equipped with linear recurrence and gating mechanisms) learned to categorize more efficiently than feedforward networks. Furthermore, networks with linear recurrence and multi-timescale gating could learn internal representations that "un-mixed" quickly-varying and slowly-varying data sources. Together, these findings demonstrate how a fundamental property of cortical dynamics (their temporal autocorrelation) can serve as an inductive bias, leading to more efficient category learning and to the representational separation of fast and slow sources in the environment.

Fanjun Bu, Chien-Ming Huang

As robots perform manipulation tasks and interact with objects, it is probable that they accidentally drop objects (e.g., due to an inadequate grasp of an unfamiliar object) that subsequently bounce out of their visual fields. To enable robots to recover from such errors, we draw upon the concept of object permanence-objects remain in existence even when they are not being sensed (e.g., seen) directly. In particular, we developed a multimodal neural network model-using a partial, observed bounce trajectory and the audio resulting from drop impact as its inputs-to predict the full bounce trajectory and the end location of a dropped object. We empirically show that: 1) our multimodal method predicted end locations close in proximity (i.e., within the visual field of the robot's wrist camera) to the actual locations and 2) the robot was able to retrieve dropped objects by applying minimal vision-based pick-up adjustments. Additionally, we show that our method outperformed five comparison baselines in retrieving dropped objects. Our results contribute to enabling object permanence for robots and error recovery from object drops.