Guangzhou Rescue Robot Contest Tests 29 Teams

Robot dogs, drones and underwater robots from 29 teams were put through mock high-rise fires and flood rescues in Guangzhou on 23 and 24 October, in the final of a city-backed contest to test emergency robots in realistic conditions. The two-day event, the “Technology for Good, Life First” Guangzhou 2026 Safety and Emergency Rescue Robot Competition, ran two tracks and seven events after about three months of trials, China News Service reported.

Alongside the results, the organisers published a method for building repeatable rescue test tracks and a plan for an open simulation model of the courses. These were demonstrations on purpose-built tracks, not operational deployments, but the organisers say strong performers will be put forward for a regional firefighting drill.

In brief

29 teams competed in a Guangzhou final with two tracks, high-rise fire rescue and natural disaster response, across six land events and one underwater event.
Xuanji Dongli won the fire track and Embodied Intelligence’s Tie’ao team won the disaster track.
Organisers released a reusable test-track design method and plan to put the seven courses into an open simulation model.

Robots were scored on search, climbing, breaching and water rescue

The competition’s organising committee says the contest is jointly run by the Emergency Management University, the Guangdong Academicians’ Federation, Guangzhou City Investment Group and the Shenzhen Public Welfare Rescue Volunteers Federation. The high-rise fire track covered finding the fire source and trapped people, carrying loads up stairs, and breaking windows to vent smoke and support firefighting from outside. The disaster track covered supply drops in mountain terrain, breaching narrow passages in collapsed buildings, underwater location and buoyancy rescue, and wide-area AI reconnaissance with precise air drops.

According to China News Service, reconnaissance robot dogs located trapped people in smoke, and four-legged robots carried heavy loads up stairs and through water. Drones flew in strong high-altitude wind to reach the outside of buildings 30 to 50 storeys high, break glass and set up water cannon. Breaching robots crawled into narrow pipes, AI drones picked out targets and dropped supplies, and an underwater robot completed a buoyancy rescue of a 160-jin (80 kg) person in murky water.

Hong Kong’s Wen Wei Po described a tracked breaching robot from Shenzhen firm Hanmeihao cutting through wooden boards, steel mesh and a reinforced concrete slab inside three 80 cm pipes. It finished the concrete stage in 26 minutes, nearly half the time limit, and won a first prize. A student team from the Hong Kong University of Science and Technology (Guangzhou), whose robot dog carried 25 kg over rubble, ditches and fallen branches, took a second prize.

Two track winners and seven first prizes

Xuanji Dongli won the high-rise fire track and Embodied Intelligence’s Tie’ao team won the natural disaster track. First prizes in individual events went to Hikvision Wuxi, Xuanji Dongli, Nanyue Emergency, Tie’ao, Hanmeihao, Qianxing Innovation and Chengfeng Yuejing, according to China News Service and Southern+.

On 24 October the organisers released two documents. One sets out a decoupled, real-scenario method for designing robot test tracks, which they call SDOT and describe as China’s first field-test scheme to cover both high-rise firefighting and natural-disaster rescue, with tracks that can be rebuilt elsewhere. The other is a plan for an emergency simulation world model that would scan and model the seven courses into a virtual space for teams to keep using.

The same day, Guangzhou Digital Intelligence Emergency Technology Group was formally launched. Led by a company under Guangzhou City Investment Group, it will focus on four areas: resilience to typhoons and rainstorms, high-rise safety and rescue, risk monitoring in industrial parks, and integrated air, space and ground hazard warning.

Organisers name the gap between demos and real rescues

At the contest’s launch in April, an engineer from the Ministry of Emergency Management’s Big Data Center, Yang Ruiyang, said that more than 90% of firefighting-robot actions still rely on manual remote control in heavy smoke and glare, and that failure rates exceed 40% in high-altitude and cold conditions, according to the organising committee. The contest’s tasks were built around those limits.

The organisers say teams and equipment that perform well and meet operational standards will be recommended by the Guangzhou Emergency Management Bureau for an advanced-equipment list for “Emergency Mission 2026”, a Greater Bay Area drill on fighting fires in super-tall buildings.

A contest score is a step short of deployment

The most useful output may be the test method rather than the trophies. Rescue robotics has long struggled with comparisons: each maker demonstrates its machine in conditions it chooses, as Unitree’s staged “guardian” promo showed this week. A shared, rebuildable track with timed tasks gives buyers and fire services a common yardstick, and an open simulation of the same courses would let teams iterate between events.

It is still a contest. Robots ran on known courses, with organisers and their partners reporting the results, and the next stage is a recommendation for a drill list, not a purchase. The organisers’ own figures on remote control and failure rates suggest how far most machines remain from working unsupervised in real smoke. That puts the Guangzhou results in the same category as Uganda’s medical-drone trial: promising evidence from controlled tests, ahead of routine service.

The open question is who buys

The launch of a city-owned emergency technology group the same day shows Guangzhou wants a local industry, not just an event. That fits China’s wider robotics push, where officials expect humanoid output to pass 100,000 units this year. In the UK, fire services are testing drones in shared airspace, as in the Chorley trial involving Lancashire Fire and Rescue.

The contest joins other projects on TSN’s Tech for Good hub. The measure to watch is whether any of the winning machines appear in the Greater Bay Area drill, and whether the SDOT track is rebuilt anywhere outside Guangzhou.