NTT DOCOMO has developed a platform that connects personal AI agents to robots and checks whether requested actions are allowed. User attributes and rules for each setting determine which instructions reach a robot. The Japanese company plans practical trials and a November demonstration, with restaurant orders among its examples of how the system could be used.
Japanese mobile operator NTT DOCOMO announced its Physical AI integration platform on October 27. It connects personal software agents, which act on users’ instructions, with different robots. The system combines robot connectivity, permission decisions and behavior adapted to nearby people. DOCOMO plans to test its practicality through internal use, joint research with robot manufacturers and trials with businesses.[1], [2]
Communication uses Agent2Agent, or A2A, an existing protocol for cooperation between software agents. DOCOMO added its own permission controls and action tracking. The central A2A Gateway compares attributes such as a user’s age with rules registered for the setting. According to the company, it halts unauthorized actions before execution and sends the robot only approved instructions. User attributes are managed within the platform rather than passed to the robot.[1]
The activity record identifies when an action happened and whose instruction initiated it. A separate behavior component generates movements in response to the robot’s surroundings, with the aim of avoiding discomfort or anxiety for people nearby.[1]
In DOCOMO’s restaurant example, a customer orders through a personal agent and the platform chooses a serving robot. Rules could prevent alcohol reaching a minor or an allergen reaching an affected diner, while movement control could steer around someone entering the robot’s path. These are proposed uses. A demonstration is planned at the NTT research forum beginning November 18 at the Musashino research center in Japan.[1]
