Avride integrates cloud VLMs as safety net for delivery robots

Avride Inc. has integrated cloud-based vision-language models (VLMs) into its delivery robots to add a proactive layer of environmental awareness. The company’s sidewalk robots already navigate busy city streets autonomously using onboard sensors and local neural networks to detect agents like cyclists, children, wheelchairs, and emergency vehicles. However, certain scenarios require deeper contextual understanding—for example, distinguishing a police officer walking home from an active crime scene. The cloud VLM acts as an automated watcher, providing holistic scene interpretation beyond basic object detection to reduce the likelihood of inappropriate behavior in unusual or high-stakes situations.

Source: The Robot Report