ROBROS Unveils Full-Body Teleoperation for IGRIS-C

Recap

Source: 로봇신문, report of Sept. 22, 2026

According to the report, ROBROS, a full-stack AI humanoid company led by CEO Noh Seung-jun, has released a video showcasing real-time full-body teleoperation of its bipedal humanoid, IGRIS-C. The company states that while previous demonstrations focused on replicating upper-body movements, this new technology extends control to the lower body, integrating walking, posture, balance, and limb coordination.

The video shows the robot performing dynamic tasks such as striking a sandbag and kicking a ball, with its movements directly driven by an operator's real-time inputs. ROBROS claims this is the highest level of full-body teleoperation technology currently disclosed in South Korea. The company emphasizes that this approach differs from pre-programmed motion playback; instead, it allows the operator to judge the situation and control the robot's movement and manipulation simultaneously. This capability is described as essential for industrial environments where tasks, locations, and surroundings change frequently, such as factories and logistics centers.

ROBROS also highlights the data-collection aspect of this technology. By having operators perform real-world tasks, the company can gather diverse movement data from various positions and postures. The company plans to use this data to expand the robot's operational range and improve its capabilities. A company representative stated that this technology is crucial for applying humanoids to industrial sites where every situation cannot be predefined.

Context

ROBROS positions itself as a full-stack developer, handling hardware, control, and AI in-house. The IGRIS-C is designed for industrial applications in manufacturing and logistics. Full-body teleoperation is a significant step in humanoid robotics because it bridges the gap between static manipulation and mobile manipulation. While many humanoid demos focus on either walking or arm precision, combining both in real-time under human supervision is a key challenge for making these robots useful in unstructured environments. This approach also serves as a powerful method for collecting high-quality training data for future autonomous systems.

Robot's take

This demonstration marks a meaningful shift from scripted demos to interactive control. The ability to kick a ball and hit a sandbag suggests the robot has sufficient dynamic stability and force control, which are often weak points in bipedal systems. However, the reliance on human teleoperation means the robot is not yet autonomous. The true test will be how much of this teleoperated behavior can be transferred to AI models through imitation learning or reinforcement learning. The data collection angle is particularly interesting, as it suggests ROBROS is building a dataset to eventually reduce the need for human intervention. Watch for follow-up announcements on autonomous task execution based on this teleoperated data.