IIIT-H and Ati Robotics Develop AI Robot Navigation Without LiDAR Maps

Hyderabad, India: Researchers at the International Institute of Information Technology Hyderabad (IIIT-H), working with Bengaluru-based Ati Robotics, have demonstrated a new approach to autonomous robot navigation using camera images, visual landmarks and AI-based 3D reconstruction.

The demonstration marks an important step in translating robotics research into an industrial environment. Instead of relying on a conventional LiDAR occupancy grid and a detailed metric map during the demonstration, the robot used visual information and an AI model to understand its surroundings and navigate towards a target on a factory floor.

The research could contribute to the development of more flexible navigation systems for autonomous mobile robots (AMRs), which are widely used in warehouses, factories and logistics facilities.

What Is the IIIT-H and Ati Robotics Breakthrough?

The Robotics Research Center (RRC) at IIIT-H has been collaborating with Ati Robotics on vision-based robot navigation since 2023. The research explores whether robots can navigate through an environment using visual cues and spatial relationships rather than depending entirely on detailed, pre-built maps.

In a recent factory-floor demonstration, an autonomous mobile robot travelled towards a specified object using a camera, images captured during an earlier walkthrough and an AI-based model that reconstructed three-dimensional information from image pairs.

The significance of the demonstration was not simply that the robot reached its destination. It was the way the robot used visual information to estimate spatial relationships and plan its movement.

How Does the AI Robot Navigation System Work?

Traditional autonomous mobile robots often use sensors such as LiDAR to measure distances, map their surroundings and estimate their positions. These systems may use simultaneous localisation and mapping (SLAM) to build a map while tracking the robot's location.

The IIIT-H research explores an alternative approach based on visual information, AI models and a more compact representation of the environment.

1. Camera-Based Visual Recognition

The robot uses camera images to identify recognisable features in its surroundings. These visual cues help the navigation system understand the environment without representing every part of the space in a dense map.

2. Sparse Localisation

Instead of relying exclusively on detailed geometric maps, sparse localisation uses a limited set of recognisable landmarks to estimate where a robot is located.

In a warehouse, these landmarks might include a doorway, an elevator or a particular section of a storage area. This is conceptually similar to how people recognise familiar places to understand where they are.

3. AI-Based 3D Reconstruction

The system can compare images of the same environment to identify corresponding visual points and estimate their spatial relationships. This information helps the robot reason about possible directions and routes.

4. Image-Based Navigation

The researchers also explored a navigation method in which an operator identifies a destination by pointing to a pixel in an image rather than entering precise three-dimensional coordinates.

For example, a target might be a warehouse pallet or a docking location. The navigation system uses visual information and estimated geometry to assess possible routes towards the selected target.

What Is LiDAR, and Why Does This Research Matter?

LiDAR stands for Light Detection and Ranging. It uses laser light to measure distances and gather information about objects and surfaces in an environment. It is commonly used in robotics, autonomous navigation and industrial mapping.

Although LiDAR can provide valuable spatial information, systems that rely on detailed maps may require additional computing resources, storage and mapping infrastructure, depending on their design and operating conditions.

Vision-based navigation investigates whether cameras and AI can perform some navigation tasks with less dependence on dense maps. If successfully developed for suitable environments, such approaches could offer additional flexibility for certain industrial applications.

Important clarification: The demonstration does not establish that LiDAR is obsolete or that all LiDAR sensors can be removed from industrial robots. It shows that a robot can navigate towards a target using an alternative, vision-based approach under the demonstrated conditions.

Has the Technology Reached the Factory Floor?

Yes, the research has moved beyond academic investigation into industrial testing. According to IIIT-H's official research blog, the institute transferred code to Ati Robotics for testing and deployment in a warehouse-robot environment.

The company has also adapted parts of the research into OccupancyReact, a LiDAR-based variant that is closer to its existing deployed solutions. This distinction matters: the collaboration is not evidence that Ati Robotics has replaced LiDAR across its entire robot fleet.

The work illustrates how academic research can be adapted by an industrial partner, with individual components retained, modified or integrated into existing systems.

Why Is the Last Two Metres Important for Warehouse Robots?

Ati Robotics has highlighted the final stage of a robot's journey as an important area for further improvement. This can involve approaching a pallet, docking at a station or moving through a narrow bay.

Even if a robot travels efficiently across a warehouse, its final approach may require careful positioning and manoeuvring. Small time savings during repeated operations could accumulate over an entire work shift.

Improving this stage could help industrial robots perform material-handling tasks more efficiently, although the actual benefits depend on the robot, workplace layout and operating conditions.

Potential Applications of AI-Powered Robot Navigation

Industry	Potential Application	Possible Benefit
Warehousing	Moving goods and approaching storage locations	More flexible navigation in suitable environments
Manufacturing	Transporting materials between workstations	Improved integration of autonomous mobile robots
Logistics	Approaching pallets and loading areas	More precise final-stage manoeuvring
Industrial facilities	Navigating around recognised landmarks	Potentially reduced dependence on dense maps for selected tasks

These are potential application areas for vision-based navigation, not a claim that every listed use case has already been validated by this particular demonstration.

Will AI Navigation Make Warehouse Robots Cheaper?

Reducing dependence on certain mapping systems could potentially lower some infrastructure or computing requirements. However, the overall cost depends on camera hardware, computing systems, software development, integration, maintenance and safety requirements.

No verified cost reduction figure was provided in the sources reviewed for this report. Therefore, it would be premature to claim that the technology will make all warehouse robots cheaper or eliminate the need for specialised sensors.

What Happens Next?

The next stage is to determine how effectively vision-based navigation performs across different industrial environments and operating conditions.

Key questions include how well the system handles changes in lighting, moving obstacles, crowded warehouse floors and environments that differ from the original images. Reliable industrial deployment also requires appropriate safety measures and testing.

IIIT-H's research and its collaboration with Ati Robotics provide a practical example of how academic work in computer vision and AI can contribute to real-world robotics systems. Further testing will help establish where the approach provides the greatest value.

Conclusion

The AI robot navigation work developed by IIIT-H and Ati Robotics demonstrates a promising direction for warehouse automation. By combining camera images, visual landmarks and AI-based 3D reconstruction, the researchers showed that a robot could navigate towards a target without relying on a conventional LiDAR occupancy grid and detailed metric map during the demonstration.

The technology does not mean LiDAR is no longer useful. Instead, it highlights how vision-based AI can complement existing robotics approaches and potentially make navigation more flexible for selected industrial tasks.

As testing and deployment continue, the collaboration will be worth watching for its potential impact on autonomous mobile robots, industrial logistics and Physical AI.

Frequently Asked Questions (FAQs)

1. What did IIIT-H and Ati Robotics demonstrate?

They demonstrated an autonomous mobile robot navigating towards a target on a factory floor using camera images and AI-based 3D reconstruction rather than relying on a conventional LiDAR occupancy grid during that demonstration.

2. What is vision-based robot navigation?

Vision-based robot navigation uses camera images and visual information to help a robot recognise its surroundings, estimate spatial relationships and determine how to reach a destination.

3. Does this technology completely replace LiDAR?

No. The demonstration showed an alternative navigation approach for a particular task. Ati Robotics has also adapted parts of the research into OccupancyReact, a LiDAR-based variant closer to its existing deployed solutions.

4. What is an autonomous mobile robot (AMR)?

An autonomous mobile robot is a robot that can move through an environment and perform tasks with automated navigation. AMRs are commonly used in warehouses, factories and logistics operations.

5. When did the IIIT-H and Ati Robotics collaboration begin?

According to IIIT-H's official research blog, the collaboration began in 2023.

6. Can vision-based navigation reduce warehouse automation costs?

It may reduce some mapping or computing requirements in suitable applications, but the overall cost impact has not been established by the demonstration alone.

7. Where can readers find the original research announcement?

Readers can consult the official IIIT-H Robotics Research Center announcement for details about the collaboration and navigation approach.

Sources and Further Reading

• IIIT-H Official Blog: Robotics Research Moves to the Shop Floor
• Deccan Chronicle: AI Vision-Based Robot Navigation Demonstration
• ET Electronics: AI Navigation for Warehouse Robots

Editorial note: This article is based on the institute's published research announcement and media reports available as of October 28, 2026. Potential applications are distinguished from results demonstrated in the reported factory-floor test.