Most fatal road accidents in urban areas involve vulnerable road users. New solutions for fighting against these accidents can be considered by leveraging connected, intelligent vehicles and smart cities connecting all parts of an urban environment. This work proposes a multi-sensing and communication approach to prevent potential accidents between vehicles and VRUs, by predicting and notifying both about potential collisions before they happen. This approach leverages and aggregates information from smart city sensors, dispersed in the vehicles (and aggregated by the On-Board Units, OBUs), in the VRUs (e.g., smartphones and smartwatches), and on the road itself (e.g. video cameras, radars, lidars). These elements communicate through several message standards and wireless access technologies (e.g. ITS-G5, C-V2X, LTE, 5G and, in the future, 6G). Using both sensing and communication, fusion and collision detection algorithms, this system predicts and notifies potential hazardous situations involving Vulnerable Road User (VRU)s and vehicles. The results in a real scenario with sensors, VRUs and vehicles on the road show that the system predicts potential collisions with high accuracy and low delay. Results also point to some vital deployment decisions that must be made to ensure proper notification timings, such as the usage of multi-homing, 5G and edge computing.