Machine learning models are increasingly being deployed in real-world contexts. However, systematic studies on their transferability to specific and critical applications are underrepresented in the research literature. An important example is visual anomaly detection (VAD) for robotic power line inspection. While existing VAD methods perform well in controlled environments, real-world scenarios present diverse and unexpected anomalies that current datasets fail to capture. To address this gap, we introduce CableInspect-AD, a high-quality, publicly available dataset created and annotated by domain experts from Hydro-Québec, a Canadian public utility. This dataset includes high-resolution images with challenging real-world anomalies, covering defects with varying severity levels. To address the challenges of collecting diverse anomalous and nominal examples for setting a detection threshold, we propose an enhancement to the celebrated PatchCore algorithm. This enhancement enables its use in scenarios with limited labeled data. We also present a comprehensive evaluation protocol based on cross-validation to assess models' performances. We evaluate our Enhanced-PatchCore for few-shot and many-shot detection, and Vision-Language Models for zero-shot detection. While promising, these models struggle to detect all anomalies, highlighting the dataset's value as a challenging benchmark for the broader research community. Project page: https://mila-iqia.github.io/cableinspect-ad/.

In this paper we present the approach for autonomous planetary exploration developed at the Canadian Space Agency. The goal of this work is to autonomously navigate to remote locations, well beyond the sensing horizon of the rover, with minimal interaction with a human operator. We employ LIDAR range sensors due to their accuracy, long range and robustness in the harsh lighting conditions of space. Irregular triangular meshes (ITMs) are used for representing the environment providing an accurate yet compact spatial representation. In this paper a novel path-planning technique through the ITM is introduced, which guides the rover through flatter terrain and safely away from obstacles. Experiments performed in CSA's Mars emulation terrain that validate our approach are also presented.

The success of NASA's Mars exploration rovers has demonstrated the important benefits that mobility adds to planetary exploration. Very soon, mission requirements will impose that planetary exploration rovers drive over-the-horizon in a single command cycle. This require an evolution of the methods and technologies currently used. This paper presents experimental validation of our over-the-horizon autonomous planetary navigation. We present our approach to 3D terrain reconstruction from large sparse range data sets, localization and autonomous navigation in a Mars-like terrain. Our approach is based on on-line acquisition of range scans, map construction from these scans, path planning and navigation using the map. An autonomy engine supervises the whole process ensuring the safe navigation of the planetary rover. The outdoor experimental results demonstrate the effectiveness of the reconstructed terrain model for rover localization, path planning and motion execution scenario as well as the autonomy capability of our approach.