Robot Fleet Cut Hospital Delivery Time 32% to 36% in a Six Month Trial

Hospital runners carried the load, and the clock ran on every leg

Inside a tertiary hospital in China, medicines, specimens and clinical supplies move all day. Pharmacy staff prepare the items, and porters carry them along busy corridors to wards and the laboratory. Each leg is a small task, but together they decide how quickly a patient receives treatment or a result.

Manual delivery is exposed to the ordinary pressures of a working hospital. Staff are pulled toward other duties, routes stretch as the day fills, and one late run can hold up a nurse or a test. The hospital's question was practical: could a robot fleet keep these runs steady and accurate, and how would it compare with a dedicated manual team?

• Pharmacy to ward medicine runs
• Ward to laboratory specimen runs
• Waiting time that builds up on every leg of a run

Ten robots tested alongside a team of 19 manual workers

The hospital deployed a fleet of ten autonomous mobile robots and ran it alongside a manual team of 19 workers. The comparison was a six month study with a parallel control design, so both methods handled delivery work over the same period and could be measured side by side.

Each robot plans its full route with Dijkstra's algorithm, then switches to A* combined with the Dynamic Window Approach for local navigation around people and obstacles. A queuing model based on M/M/c theory coordinates how the ten units share incoming work, so jobs are spread across the fleet rather than piled onto one machine.

• Global route planning with Dijkstra's algorithm
• Local navigation with A* and the Dynamic Window Approach
• Fleet coordination through an M/M/c queuing model
• Parallel comparison against manual delivery over six months

Faster legs, more trips and verified items

Across the study, robot delivery cut delivery time by 32% to 36% compared with manual delivery, a difference reported as statistically significant (p < 0.001). The result held across the six months rather than appearing in a single good week.

The 10 robot fleet completed 7.3 times more delivery trips than the 19 manual workers over the same six months. On individual legs, pharmacy to ward runs averaged 21 minutes by robot against 31 minutes by hand, and ward to laboratory runs averaged 23 minutes against 38 minutes. In June 2025 the fleet logged 10,500 delivery trips with an average trip time of 23.9 minutes and system uptime of 94.2%, so the team still had to plan around downtime.

Accuracy is where the robot group stood out most clearly. It reached 100% verification accuracy and 100% item integrity, while manual delivery recorded 97% and 99%. The robot group also had a significantly lower omission rate (p = 0.011).

Nurses and patients both rated the service more favorably in the robot group, with significant differences (p < 0.05). Over a 10 year lifecycle, the study estimates the robot system saved 6.928 million RMB, and a sensitivity analysis was included to test that estimate.

What a fleet decision looks like for a US operator

The study gives a hospital operator a measured picture of how a robot fleet and a manual team compare on time, trips and accuracy. Service Robot Co. is a full service, OEM neutral commercial robot integrator for US businesses. We select the right robots across manufacturers, then finance, deploy, integrate, train and service every unit through a nationwide US engineer network, so one partner covers the whole lifecycle.

Customers get our robots through a lease, not a sale. Service Robot Co. leases robots only, on a lease running three years with one fixed payment quoted after a site assessment. That assessment maps routes, floor conditions, elevator access and staffing patterns, which are the same questions that decide fleet size. Monthly payment programs can replace a large upfront commitment while a team tests the numbers on its own floors.