Verification tests began during 1999. SCB used actual machines to clean real buildings. "Our role was to uncover areas for improvement from the user's perspective, and improve performance to a commercially viable level." (Miyata)
As opposed to vacuum cleaners in private households, the robots would have to be able to clean all the dirt the first time around. However, the team discovered that moving at a set speed resulted in uneven cleaning performance. In response, the team developed a cleaning system with three separate speed settings depending on the amount of dirt present.
In response to requests from SCB, Fuji Heavy Industries brought its technical capacity to bear, providing technology to allow the robot to move in a straight line over carpeted services, and a mechanism to avoid clogs.
"The people at Fuji Heavy Industries were quick to respond to even the most detailed request. Given their enthusiasm, I was positive that the project would succeed." (Ishikawa)
The robots do not clean within 20 cm of a wall. While technologically feasible, the number of sensors required to accomplish this task would dramatically increase, adding significantly to the cost of the system. "Humans don't walk right next to walls. That's why we decided on a cleaning policy that any noticeable dirt next to a wall can be cleaned by a human worker." (Izawa)
These experiments also confirmed that this separation of responsibilities made for an effective overall system. And why 20 cm? "The nozzle width of most vacuum cleaners is 20 cm." (Miyata)
As verification tests entered the final stages, the team thought they were close to a commercial product. Then, the team discovered what was perhaps their biggest problem to that point: carpet tracks. With a 135kg robot traveling the same path every day, the carpet was sure to wear out. "We were only able to uncover the problem because we tested in an actual building." (Izawa)
The team tested numerous types of carpets with the cooperation of Sumisho Interior International, but the results were all the same. The team made the robot's wheels wider to improve the situation, but the measure didn't satisfy the team. Lighter batteries were considered, but rejected due to cost issues.
The team decided to resolve the problem by having the robots take different paths each day. "We figured that altering the path a few centimeters on different days would make carpet ruts less likely." (Miyata)
Having introduced this last tweak, the system was completed in December 2000. The dreams and desires of Inoue and the team became a real system that was put into operation in April 2001, the same day as the Harumi Triton grand opening.
One major factor behind the success of this project is that it was not conceived as a robot, but as a building cleaning system that happened to incorporate robots. In other words, in addition to robots and elevators, the team considered how to combine people performing complementary cleaning, people operating/ managing the robots, and people using the building with the robot/elevator system to achieve overall quality and cost balance.
The Sumitomo Corporation Group provided important technologies allowing for the maximum usage of Fuji Heavy Industries monodukuri technologies. This combination led to the commercialization and implementation of an advanced service robot system. (Hajime Aoyama/ engineering professor; Fuji Heavy Industries Clean Robot Department Manager)
