LiDAR in Warehouse Automation: A Real-World Look at the RoboSense Helios 32

Warehouse automation is one of the fastest growing application areas in mobile robotics right now. Autonomous Mobile Robots (AMRs) are being deployed across large-scale logistics and fulfillment environments to handle everything from localization and navigation to obstacle avoidance and dynamic mapping. Often operating alongside forklifts, moving pallets, and human workers in real time.

Getting that right is hard. And one of the most critical components in the stack, one that often makes or breaks how reliably an AMR can actually do its job, is the LiDAR.

We've been running the RoboSense Helios 32 on a warehouse AMR, and here's what we found.

 

 

THE SETUP

Our AMR operates in a large-scale warehouse environment; wide aisles, repetitive visual features, metal racking, forklifts, and constantly shifting pallet positions. It's exactly the kind of environment that stress-tests every sensor in the stack.

The Helios 32 is front-mounted on the robot. That's a constraint driven by the robot's design, the top deck is reserved for other use, and it's worth noting because it put the sensor in a more exposed position than a traditional top-mounted configuration. As it turns out, that became an accidental durability test. More on that shortly.

Our autonomy stack pairs the Helios with a ZED X stereo camera, using RTAB-Map for SLAM. The camera handles visual estimation; the Helios handles geometric precision. Together, they form a sensor fusion architecture that's proven more robust than either sensor alone.

 

WHAT WE USE IT FOR

Geometric Grounding and Localization

In a warehouse, visual SLAM has a well-known weakness: repetitive environments. When every aisle looks the same; same racking, same lighting, same colour palette. Camera-based localization can drift or lose confidence, particularly in long straight runs between identical shelving bays.

The Helios 32 solves that. Its 32-channel point cloud provides the precise geometric data needed to anchor the robot's position within the warehouse map, even in areas where the visual features are ambiguous or repetitive. The LiDAR doesn't care that aisle 7 looks like aisle 12. It reads the geometry; the exact distances, angles, and structural features, and that's enough to maintain a reliable position fix.

The result is a localization pipeline where the cameras provide a strong initial visual estimate, and the Helios provides the geometric certainty to confirm and correct it. That combination has proven significantly more stable than relying on either sensor independently.

Obstacle Detection and Safety

Beyond mapping and localization, we've been developing obstacle detection using the Helios's point cloud data. A 32-channel sensor at this resolution can detect floor-level hazards; a box left in an aisle, a fallen object or a pallet slightly out of position. A camera system may miss that, either due to viewing angle, lighting conditions, or simply the inherent difficulty of resolving small ground-level objects through computer vision alone.

This work is still in active development, but the early results are promising. The low-noise indoor performance of the Helios 32 ensures very little unwanted reflection from metal shelving surfaces, which can be a real problem in warehouse environments. This means that the point cloud data is clean enough to work with without significant pre-processing overhead.

The 32-channel configuration is worth addressing directly here, because the market does offer 64 and 128-channel alternatives. For many warehouse applications, 32 channels is genuinely sufficient. The repeating scan pattern means very small objects that fall exactly between scan lines can theoretically be missed, but the angular resolution of the Helios 32 makes that a rare edge case rather than a practical concern. In our deployment, it hasn't been an issue.

Sensor Fusion

The longer-term architecture we're building toward combines the Helios's depth accuracy at range with the ZED X's rich visual data to create a perception layer that handles dynamic environments robustly. Warehouses are not static. Pallet positions shift, traffic patterns change, temporary blockages appear and disappear. A sensor fusion approach, geometric + visual, is the right way to handle that, and the Helios's reliable, consistent depth data is the anchor that makes it work.

 

WHAT STOOD OUT

Setup Was Surprisingly Simple

The Helios 32 connects to a self-hosted web interface via its device IP. Parameter configuration is done through that interface. This simple setup requires no complex software toolchain, no lengthy initialization process. For engineers who've wrestled with LiDAR setup before, this is a genuine breath of fresh air. You spend your time integrating and building, not fighting the sensor.

The ROS SDK Has a Useful Trick

One feature worth calling out: the Helios ROS SDK allows you to filter out data points from regions obstructed by the robot itself. For a front-mounted sensor, part of the field of view inevitably captures the robot's own body. Without filtering, that generates spurious obstacles in your system. The SDK's built-in filtering handles this cleanly, which is a small but meaningful quality-of-life feature for anyone working on a custom mounting configuration.

The Range and Signal Quality Are Solid

Indoor performance has been excellent. Clean point clouds, consistent range data and low noise from metal surfaces. Warehouse racking and shelving can cause significant interference for some sensors. The Helios 32 handles the metal environment well, maintaining reliable returns across multiple aisles without the clutter that would otherwise require additional filtering work.

THE DURABILITY STORY

This one deserves its own section, because it's the kind of real-world proof point that no spec sheet can give you.

During one of our deployments, a forklift collided with the front of the robot while it was at its charging station. The impact was significant enough to bend the sensor mount and flip the robot. The Helios 32, sitting at the front directly in the collision zone, survived! It is still in active use today.

That kind of physical resilience matters in a live warehouse environment. Forklifts and AMRs sharing space is a reality, and collisions, however rare, can happen. A sensor that survives a major impact and keeps working is not just convenient, it shows how well the sensor is built.

BOTTOM LINE

The RoboSense Helios 32 has proven itself as a reliable, capable, and durable LiDAR for warehouse AMR applications. It's not the highest channel count on the market, but for most warehouse navigation, localization, and obstacle detection use cases, 32 channels at this resolution is the right tool for the job, and it comes with setup simplicity, clean indoor performance, and clearly excellent build quality.

If you're building or evaluating an AMR for warehouse deployment and weighing your LiDAR options, the Helios 32 is worth a serious look.

The RoboSense Helios 32 is available on the InDro Robotics Store. Browse the full LiDAR collection at store.indrorobotics.com/collections/lidar or reach us directly at sales@indrorobotics.com to discuss your application.