Nowadays, unmanned aerial vehicles (UAVs) are commonly used in search and rescue scenarios to gather information in the search area. The automatic identification of the person searched for in aerial footage could increase the autonomy of such systems, reduce the search time, and thus increase the missed person's chances of survival. In this paper, we present a novel approach to perform semantically conditioned open vocabulary object tracking that is specifically designed to cope with the limitations of UAV hardware. Our approach has several advantages: It can run with verbal descriptions of the missing person, e.g., the color of the shirt, it does not require dedicated training to execute the mission, and can efficiently track a potentially moving person. Our experimental results demonstrate the versatility and efficacy of our approach. We publish the methods source code at https://github.com/utn-blei/CloudTrack.

In this paper, we propose VLM-Vac, a novel framework designed to enhance the autonomy of smart robot vacuum cleaners. Our approach integrates the zero-shot object detection capabilities of a Vision-Language Model (VLM) with a Knowledge Distillation (KD) strategy. By leveraging the VLM, the robot can categorize objects into actionable classes-either to avoid or to suck-across diverse backgrounds. However, frequently querying the VLM is computationally expensive and impractical for real-world deployment. To address this issue, we implement a KD process that gradually transfers the essential knowledge of the VLM to a smaller, more efficient model. Our real-world experiments demonstrate that this smaller model progressively learns from the VLM and requires significantly fewer queries over time. Additionally, we tackle the challenge of continual learning in dynamic home environments by exploiting a novel experience replay method based on languageguided sampling. Our results show that this approach not only reduces energy consumption by 53 % compared to cumulative learning but also surpasses conventional vision-based clustering methods, particularly in detecting small objects across diverse backgrounds.

Effective robot navigation in unseen environments is a challenging task that requires precise control actions at high frequencies. Recent advances have framed it as an image-goal-conditioned control problem, where the robot generates navigation actions using frontal RGB images. Current state-of-the-art methods in this area use diffusion policies to generate these control actions. Despite their promising results, these models are computationally expensive and suffer from weak perception. To address these limitations, we present FlowNav, a novel approach that uses a combination of Conditional Flow Matching (CFM) and depth priors from off-the-shelf foundation models to learn action policies for robot navigation. FlowNav is significantly more accurate and faster at navigation and exploration than state-of-the-art methods. We validate our contributions using real robot experiments in multiple environments, demonstrating improved navigation reliability and accuracy. Code and trained models are publicly available†.

Visual place recognition is essential for vision-based robot localization and SLAM. Despite the tremendous progress made in recent years, place recognition in changing environments remains challenging. A promising approach to cope with appearance variations is to leverage high-level semantic features like objects or place categories. In this paper, we propose FM-Loc which is a novel image-based localization approach based on Foundation Models. Our approach uses the Large Language Model GPT-3 in combination with the Visual-Language Model CLIP to construct a semantic image descriptor that is robust to severe changes in scene geometry and camera viewpoint. We deploy CLIP to detect objects in an image, GPT-3 to suggest potential room labels based on the detected objects, and CLIP again to propose the most likely location label. The object labels and the scene label constitute an image descriptor that we use to calculate a similarity score between the query and database images. We validate our approach on real-world data that exhibit significant changes in camera viewpoints and object placement between the database and query trajectories. The experimental results demonstrate that our method is applicable to a wide range of indoor scenarios without the need for training or fine-tuning.

Tracking the pose of a camera is at the core of visual localization methods used in many applications. As the observations of a camera are inherently affected by lighting, it has always been a challenge for these methods to cope with varying lighting conditions. Thus far, this issue has mainly been approached with the intent to increase robustness by choosing lighting invariant map representations. In contrast, our work aims at explicitly exploiting lighting effects for camera tracking. To achieve this, we propose a lighting adaptable map representation for indoor environments that allows real-time rendering of the scene illuminated by an arbitrary subset of the lamps contained in the model. Our method for estimating the light setting from the current camera observation enables us to adapt the model according to the lighting conditions present in the scene. As a result, lighting effects like cast shadows do no longer act as disturbances that demand robustness but rather as beneficial features when matching observations against the map. We leverage these capabilities in a direct dense camera tracking approach and demonstrate its performance in realworld experiments in scenes with varying lighting conditions.

The ability to build models of the environment is an essential prerequisite for many robotic applications. In recent years, mapping of dense surface geometry using RGB-D cameras has seen extensive progress. Many approaches build colored models, typically directly using the intensity values provided by the camera. Unfortunately, these intensities are inherently affected by illumination. Therefore, the resulting maps only represent the environment for one specific lighting condition. To overcome this limitation, we propose to build reflectance maps that are invariant against changes in lighting. Our approach estimates the diffuse reflectance of a surface by recovering its radiosity and the corresponding irradiance. As imperfections in this process can significantly degrade the reflectance estimate, we remove outliers in the high dynamic range radiosity estimation and propose a method to refine the reflectance estimate. Our system implements the whole pipeline for offline reconstruction of dense reflectance maps including the segmentation of light emitters in the scene. We demonstrate the applicability of our approach in real-world experiments under varying lighting conditions.