PAL Robotics | Demo at VivaTech 2026

In this talk

1. 0:01

   A decade of research and industry collaboration

   Albert introduces a demonstration of the Kangaroo and a humanoid robot. Olivier says PAL and PAL France have worked with his team for 10 years, developing humanoids for industrial tasks that require strong arms and legs. He describes early work on motion and torque control, including a European project that extended techniques used with Boston Dynamics’ Atlas.

2. 2:08

   From mobile manipulation to industrial deburring

   Olivier describes torque-controlled robots that can be physically interacted with, then introduces the I-GAMUS project and a lower-cost mobile manipulator. The system handles planning and real-time torque control for deburring, including work inside complex Airbus components. He says the joint lab with PAL Robotics takes many feedback cycles to develop new robots.

3. 3:18

   TIAGo Pro for safer object handling

   Albert presents TIAGo Pro, designed for moving objects between locations in industrial settings. Its two seven-degree-of-freedom arms use torque control developed with LAAS through AgileMoose, and automatic brakes hold the arms in place during an emergency stop. LiDARs detect nearby people, while cameras support SLAM navigation; Albert describes these as safety layers for certification in specific applications.

4. 5:41

   ROS 2, research access, and human interaction

   Albert says PAL robots use ROS 2 and that PAL maintains open-source repositories so users can build applications or conduct research. He also describes the company’s focus on social robotics and human-centered deployment, including virtual eye modules for interaction. TIAGo Pro is presented for both research and industrial use.

5. 7:35

   Humanoid actuation, payload, and reinforcement learning

   During the demonstration, a presenter describes a differential-motor actuation system that heats less than the compared design and says the robot can carry up to 60 kilograms. The T-REX project, funded by France 2030, will test heavier loads. The presenter says reinforcement learning produced walking results in one year, and argues the leg architecture may better withstand falls; the team is also comparing PAL robots with other companies’ systems.