Physical AI News (Japan)

Physical AI, robotics and compute — reported from Japan.

日本語
· Research & technology· 日本語で読む

PAL Robotics evaluates CoreSense for explaining robot states with TIAGo

PAL Robotics used TIAGo as the social-robot testbed for CoreSense, a four-year Horizon Europe project. Using voice and vision modules and HEXAR, the robot communicated its actions, such as assisting visitors, and its internal state through speech and gestures. The resulting ROS-compatible software is scheduled to be available and maintained from 2026 through 2029.

Key facts

  • CoreSense is a four-year Horizon Europe project. PAL Robotics led the social-robot testbed and deployed and evaluated the project on TIAGo.
  • The test scenarios involved welcoming visitors, guiding them to seats and taking care of their belongings. TIAGo used voice and vision modules and HEXAR.
  • The project’s outputs are a hybrid cognitive architecture and a ROS-compatible open-source toolbox.
  • The resulting ROS-compatible software is scheduled to be available and maintained from 2026 through 2029.
  • PAL Robotics announced that research on explainability won the Best Paper Award in the Human-Robot Interaction category at ICRA 2026.

What happened

PAL Robotics announced on November 18, 2026, that it deployed and evaluated TIAGo as a social-robot testbed in CoreSense, a four-year Horizon Europe project. The tests covered scenarios in which TIAGo welcomed visitors, guided them to seats and took care of their belongings. It activated voice and vision modules as appropriate to the situation. The company said the robot communicated its internal state through speech and gestures using HEXAR.

Why it matters

CoreSense evaluated, in visitor-assistance scenarios with TIAGo, an approach that represents robot states with explicit models and communicates them through speech and gestures using HEXAR. PAL Robotics says it will make the resulting software available and maintain it from 2026 through 2029, putting the availability of the research outputs and the effectiveness of explanations in real-world settings in focus.

Analysis

CoreSense focuses less on the robot’s capabilities and more on explicitly modeling what the robot understands and which task it is performing, then communicating that state to people. According to PAL Robotics, engineers create the models and provide them to the robot, so it does not need to be retrained for each situation. This is significant because the project evaluated cognitive and explanatory functions as part of task execution in a setting such as visitor assistance, where people need to interpret the robot’s intentions. However, judging the practical impact would also require evaluation metrics not mentioned in the article, such as the accuracy of explanations and users’ level of understanding. We previously covered PAL Robotics launches ARAL, a humanoid development project for logistics (November 4, 2026), which addressed humanoid development for logistics environments, and humanoid.guide introduces TIAGo Pro (October 30, 2026), which introduced a mobile manipulator. Unlike those product and application introductions, this announcement describes a case in which TIAGo was used as a platform to evaluate research outcomes. Whether CoreSense functions can be integrated into TIAGo and ARI remains a subject for future evaluation; the announcement does not indicate that implementation in the products has been completed. Structurally, explicit models, ROS-compatible tools, TIAGo’s voice and vision capabilities, and state explanations through HEXAR form a process for evaluating theory in an on-robot human-interaction task. The test environments also include manufacturing and drone inspection, so the results are not limited to social robots. However, the implementation details and performance comparisons for each environment are not provided, making it difficult to judge how far the shared architecture can be applied to different tasks. The plan to maintain the software from 2026 through 2029 assumes that a user community will carry the results forward after the research concludes, but the number of users and the specifics of the maintenance structure are not stated. The next points to watch are the use and updates of the released tools, the consistency of explanations in real-world settings, and the results of integration evaluations with TIAGo and ARI.

Details

According to PAL Robotics, the CoreSense consortium developed a theory of autonomous robots’ “awareness” and turned it into a hybrid cognitive architecture and a ROS-compatible open-source toolbox. The approach uses explicit models of the robot itself, its tasks and the objects it handles, created by engineers and provided to the robot. The company says this eliminates the need for retraining for each new situation. The results were tested in three environments: manufacturing, drone inspection and social robotics. PAL Robotics led the social-robot testbed. The ROS-compatible software resulting from the project is scheduled to be available and maintained from 2026 through 2029. The company will evaluate whether CoreSense’s awareness and reasoning functions can be integrated into TIAGo and ARI. The company also said that research on explainability won the Best Paper Award in the Human-Robot Interaction category at ICRA 2026.

Sources

Translated from the Japanese original: /news/pal-robotics-tiago-de-robotto-no-joutai-wo-setsumeisu-ru-cor-9b0705. The Japanese version is authoritative if they differ.