Hesai Launches Kosmo 3D Camera to Advance AI-Powered Robots

Chinese lidar technology company Hesai has introduced a new 3D spatial mapping camera designed to help developers improve the way artificial intelligence-powered robots understand and interact with the physical world.

Unveiled on October 14, 2026, the device, called Kosmo, combines lidar sensors with cameras to collect detailed spatial information. The technology is intended to help address a major challenge in robotics: providing AI systems with enough high-quality real-world data to perform useful tasks reliably.

As companies work to move humanoid robots beyond demonstrations and simple movements, spatial perception is becoming an increasingly important part of robotics development.

What Is Hesai's Kosmo Camera?

Kosmo is a spatial data collection device that combines camera technology with lidar sensing. Lidar measures distances using laser light, helping systems build a detailed understanding of objects and their positions in an environment.

When combined with visual information from cameras, spatial measurements can help developers build more detailed representations of real-world surroundings.

For AI-powered robots, this information can be valuable during training. A robot needs to recognize objects, estimate distances, understand how items relate to one another, and determine how to move safely when interacting with its environment.

The quality of this information can influence how effectively a robot learns to perform tasks outside controlled laboratory conditions.

Why Real-World Data Matters for Robotics

Training a robot to perform useful physical tasks involves more than teaching it to recognize images or follow text instructions. It must also understand spatial relationships and respond to changes in its surroundings.

For example, a robot working in a warehouse may need to identify a package, estimate its position, reach toward it, and move it without colliding with nearby objects.

A robot assisting in a retail environment may need to recognize products, navigate around people, and manipulate objects with precision.

These tasks require a combination of perception, planning, movement, and feedback. Detailed spatial data can help developers train and evaluate the systems responsible for these capabilities.

Hesai co-founder and CEO David Li said the robotics industry needs better spatial data to move beyond low-impact demonstrations and toward solving practical productivity challenges.

From Robot Demonstrations to Useful Work

Humanoid robots have attracted attention through demonstrations involving walking, dancing, waving, and interacting with people. However, reliable performance in everyday work environments is considerably more difficult.

A useful robot must handle different objects, changing lighting, unexpected obstacles, and variations in the way tasks are performed.

Improving spatial perception is one part of addressing these challenges. Developers also need reliable hardware, effective AI models, precise actuators, appropriate safety systems, and sufficient testing.

Kosmo is intended to support the data collection side of this process. Hesai said it has signed agreements with Chinese robotics companies that plan to use the device.

The company did not disclose revenue projections for the new product in its announcement.

Hesai's Broader Robotics Strategy

Hesai is known for lidar technology used in autonomous driving. The company is also developing robotic actuators, which convert electrical energy into mechanical movement.

Actuators are important components in robotic systems because they influence how accurately and efficiently a robot can move its joints and manipulate objects.

Hesai said its actuator technology is already being used in the Sharpa Wave robotic hand. The company also cited a demonstration in which a robot autonomously served ice cream at a Dairy Queen outlet in Shanghai, reporting a task success rate of 99%.

That figure relates to the reported demonstration and should not be interpreted as evidence that robots can achieve the same success rate across different tasks or operating environments.

Hesai and Sharpa Robotics operate independently, although they have connections through their founders.

What This Could Mean for Businesses

If spatial data collection becomes more efficient, robotics developers may be able to improve training workflows and test physical AI systems in a wider range of environments.

Potential applications include warehouse operations, manufacturing, retail assistance, logistics, and other settings where robots must interact with objects and people.

However, a new sensing device alone does not guarantee that a robot will become commercially useful. Businesses will still need to evaluate performance, reliability, integration costs, maintenance requirements, and safety before deploying robots at scale.

The practical value of Kosmo will depend on how well developers can use the collected data to improve robot capabilities and whether those improvements translate into dependable performance in real working conditions.

Challenges That Remain

Physical AI faces several technical challenges. Robots must operate in environments that can change unexpectedly, recognize unfamiliar objects, recover from mistakes, and avoid causing harm.

Collecting data is only one step. Developers must also ensure that the data is representative of real environments, that training methods generalize to new situations, and that the resulting systems can be tested safely.

Costs and hardware availability will also influence adoption. Businesses considering robotic automation need to compare the benefits of improved automation with the expense of equipment, integration, training, and ongoing support.

The Bigger Picture

Hesai's Kosmo launch reflects a wider shift in AI development from systems that primarily process digital information toward systems that perceive and act in the physical world.

As AI-powered robots become more capable, sensors and spatial data infrastructure may become increasingly important parts of the technology ecosystem.

The next stage of robotics will depend not only on more advanced AI models, but also on the quality of the data, sensing hardware, movement systems, and safety processes that allow those models to operate in the real world.