Joint Research on Robot Friendly Environment
(Ministry of Economy, Trade and Industry)
We conducted collaborative research towards the establishment of a robot-friendly environment. The theme of this project is to "demonstrate options for countermeasures through facilities, robots, and humans, and aim for rational robotization according to the situation."
The research was conducted as part of a project subsidized by the Ministry of Economy, Trade and Industry, with the aim of further promoting the application and spread of robot technology.
Demonstration Project Structure and Demonstration Location
The research was conducted by four companies: Tokyu Land Corporation, Tokyu Community Corporation, Softbank Corporation, and Nikken Sekkei Ltd.
SoftBank’s Chief Scientist Office was in charge of the following:
-
Robot technology
-
Delivery robot management
-
Operational issues (identification of issues related to the operation of delivery robots in shared office areas)
-
Facility countermeasures (verification of the effectiveness of some proposed countermeasures using delivery robots)
-
Robot countermeasures (verification of countermeasures for delivery robots in shared office areas)
-
Human countermeasures (verification and analysis of effectiveness in shared office areas)
Demonstration Robot
The autonomous robot Cuboid, which is being researched and developed in the Chief Scientist Office, was used for the demonstration.
Technology used in the demonstration experiment
(people flow analysis using LiDAR sensors)
As part of the demonstration experiment, we conducted pedestrian flow analysis using LiDAR sensors to investigate how various measures affect human movement in environments where humans and robots coexist. Specifically, we compared differences in trajectories for three measures: stickers, warning sounds, and wider detours. The results confirmed that warning sounds and wider detours were more effective measures.
Demonstration Experiment
(Sensor anomaly due to interior finish)
Cuboid’s depth camera experiences sensor anomaly due to reflection caused by interior finishing of the doors, resulting in Cuboid mistakenly perceiving nearby objects and taking obstacle avoidance actions.
We conducted an experiment by applying a different material to a door that had been misidentified by the depth camera (the material that was thought to be the cause of misidentification was a glossy material, so we selected a non-glossy material). The results showed that the misrecognized obstacles were not detected after the countermeasures were taken, indicating that the situation had been improved.
The false recognition in combination with a specific interior finish as shown in this case may occur in other robots as well, and this case is considered to be a useful reference for others.
Future Challenges and Prospects
Generally, robot developers/providers present the operational environment as an Operational Design Domain (ODD) to the operator, but the environment in a facility is diverse and difficult to cover, and there are situations that are difficult to foresee, such as the case of the sensor abnormality caused by the interior. It is also difficult to say that there is currently no consensus among robot operators on the standards and information to be presented. In the future, we would like to consider how to present information from the robot operator side and how to categorize the level of the robot side, using the robot friendliness level as a clue.
