- The quadruped system is designed for frontline inspections in high-risk terrain during flood season
- The launch comes as extreme weather raises pressure on local emergency response capabilities
A new AI-powered robot designed for flood control and disaster response has been introduced in eastern China, as authorities look to improve monitoring and rescue operations in hazardous environments.
The system, jointly released on April 16 by quadrupedal robot developer DEEP Robotics and Zhejiang’s Emergency Management authorities, was demonstrated during a live drill in Hangzhou’s Yuhang District, simulating geological hazard inspections under flood-season conditions.
Built on the M20 quadruped robot platform, the system integrates AI models, onboard computing and multi-sensor payloads, including thermal imaging and communication modules.
It is designed to carry out tasks such as risk detection, site reconnaissance and search operations in areas that may be inaccessible or unsafe for human responders.
Zhejiang has entered its annual flood season, when heavy rainfall and typhoons increase the likelihood of landslides, falling debris and infrastructure damage, particularly in mountainous regions.
Traditional inspection methods have struggled with limited access to high-risk zones, slower response times and inconsistent hazard detection, prompting efforts to introduce automated tools.
During the exercise, the M20 robot navigated simulated disaster scenarios such as fallen trees and unstable slopes, using AI-based detection systems to identify risks and transmit data in real time to a command platform.
Edge computing capabilities allow most processing to be handled locally, reducing reliance on cloud connectivity and improving response speed.
The robot is capable of traversing uneven terrain, climbing slopes of up to 45 degrees and operating in shallow floodwaters, enabling it to perform reconnaissance and data collection in complex environments.
The deployment reflects broader efforts to incorporate embodied AI systems into emergency management workflows, particularly at the local level where resources and access can be constrained.
