Does your vendor know your pain points?

For autonomous mobile robot (AMR) developers, creating obstacle detection based on 3D point clouds isn’t nearly as straightforward as it seems. 

If you’re reading this, you probably know this already. And you’ve probably hit the two biggest physics hurdles: floor segmentation and false positive mitigation. 

Using a third-party ODS system would be ideal. But, if you’ve looked around for one, you’ve discovered the two real challenges in solution development: 

1. Finding a vendor that understands your use case and its challenges. 

2. Finding a vendor with a purpose-driven solution for them

Here’s the good news: ifm addressed both problems at once. 

By truly understanding the physics behind the unique problem statements in obstacle detection, we developed an ODS tailor-made to take your autonomous mobile robots to the next level. 

In this article, we’ll explore those two problem statements, examine why so many vendors don’t fit the bill, and explain what we did differently – and how it will help you.

The two big ODS problem statements

Each unplanned stop your bot makes leads to increased downtime, as a human must intervene before the robot can resume its mission. 

Very quickly, that inefficiency equals big losses in productivity for your clients. 

Robust floor segmentation and false positive mitigation in your ODS significantly reduce those unplanned stops.

Floor segmentation

Navigating the diverse nature of floors, from concrete to shiny tiles, poses unique challenges for technologies such as indirect time of flight, LiDAR, and stereo cameras. 

Poor floor segmentation means the robot can’t tell the floor from an object on the ground. Or, it can’t recognize a negative obstacle like stairs or the edge of the loading dock. That leads to collisions. 

False positives

False positives occur when the system wrongly flags an obstacle. The robot stops when it struggles to differentiate between the floor and objects or environmental reflections (e.g., shiny metal or dust). 

In this case, there’s no collision. But, the robot stops when it shouldn’t, leading to downtime. 

Four reasons these problem statements are so challenging

The critical challenges to floor segmentation and false positives are:

1. Ensuring accurate recognition of all components on the same plane

2. Distance determination

3. Reflective differences between shiny and dark objects

4. Safety standards (e.g., specific reflectance targets) introduce additional noise

To tackle these challenges, you must merge diverse sensor modalities to establish a uniform ground plane for precise object detection. 

So why hasn’t anyone come up with a good solution until now?

One size rarely fits all (or any)

A common mistake in hardware development is attempting to create a one-size-fits-all solution. 

This strategy doesn’t produce a foolproof solution with a specific purpose. Instead, it often results in a product that performs adequately – at best – across multiple applications.

That’s a real problem for obstacle detection today when we can’t lower the price of an AMR much more. Instead, the best way to provide value is to make a more efficient robot with better ROI for your client. 

You need an innovative ODS that increases throughput by decreasing downtime to do it. 

That’s where we stepped in.

ODS hardware and software designed with your AMR in mind

ifm leveraged years of experience in the robotics field to develop the Obstacle Detection System using our O3R perception platform. We also prioritized understanding and addressing customers’ use cases and challenges. 

Instead of partially addressing many problems, we drilled down on what we knew were the two biggest hurdles: floor segmentation and false positive mitigation. 

The O3R is flexible and customizable. You can purchase the software and camera system or just one of them.

The programming languages are open-source (Linux) or industry standard (NVIDIA). So, there’s no getting boxed into a proprietary system or working in a language you don’t often use. 

Did we mention it works? 

With our energy and resources going into the two most challenging problem statements, the O3R addresses those pain points like virtually no other product on the market. 

Here’s a quick example of what a regular camera sees, along with how the O3R interprets the data: 

With a robust out-of-the-box ODS that integrates easily into any build, you can focus on what makes your robot unique – not figuring out how to reduce friction in its obstacle detection.

<< Faster perception stack processing | Increase your mobile robot ROI>>