The field of indoor localization is moving towards more robust and adaptive solutions that can handle dynamic and heterogeneous real-world settings. Researchers are exploring novel approaches to address the challenges of domain and class shifts, which can render static machine learning models ineffective over time. One of the key directions is the development of unified continual learning frameworks that can jointly address domain-incremental and class-incremental learning scenarios. Another area of focus is the improvement of Wi-Fi Fine Time Measurement protocols, such as 802.11mc and 802.11az, to achieve meter-level location accuracy. Additionally, there is a growing interest in reducing the overhead of fingerprinting-based indoor localization systems, with techniques such as spatial augmentation and physics-informed diffusion models showing promising results. Noteworthy papers include:

• A novel unified continual learning framework that achieves significant improvements in mean localization error and forgetting.
• A comparative study of 802.11mc and 802.11az protocols, which demonstrates the capability of 802.11az to provide better accuracy in multipath-heavy environments.
• A spatial augmentation framework that reduces fingerprinting overhead by generating high-quality synthetic data at unseen locations.
• A physics-informed diffusion model that enables accurate indoor radio map construction and localization.