wheel.me levels up navigation robustness in dynamic environments with RGo's Perception Engine

Autonomous Mobile Robots (AMRs) are revolutionizing industries with their ability to navigate and operate in complex environments without human intervention. However, one of the most significant challenges they face is localization—determining their precise location within an environment. Knowing exactly where a mobile robot is in its environment – is crucial for safe and efficient operation. This task becomes particularly challenging in a dynamic and very repetitive environment (where many aisles etc. look the same). Here are some common navigation challenges that mobile robots face:

‍

• Dynamic environments: Warehouses and factories are constantly evolving. Production lines shift, layouts changes, and unexpected obstacles appear. Traditional robots can get delocalized and inefficient in these situations.

• Limited sensor capabilities: Basic sensors like 2D LiDARs and cameras can be affected by dust, smoke, or poor lighting, leading to inaccurate readings and navigation errors. Also, 2D LiDARs rely on static objects for localization, making it hard for them to adapt to changing environments.

• Magnetic interference: Metal racks and equipment can create magnetic fields that disrupt sensors, causing robots to veer off course.

• Ambiguous landmarks: In repetitive environments like grid-like warehouse aisles or identical office corridors, distinguishing between similar-looking areas is challenging. Traditional mobile robots often struggle to identify unique features that help in determining their precise location. 

‍

To ensure robust navigation in dynamic environments, wheel.me is using RGo's Perception Engine. RGo's Perception Engine is a cutting-edge modular AI software stack designed to address common challenges faced by mobile robots in ever-changing environments. It provides real-time localization, obstacle detection, and object recognition data via an API.

‍

Here’s how it works:

The Engine provides robots with a constant stream of data through an API, giving the robots real-time awareness. This data includes:

• Precise and robust localization: The engine uses 3D vision-based advanced SLAM, AI algorithms, and fuses multiple modalities (f.e. smart cameras) to pinpoint exactly where the robot is. This is crucial for safe and efficient navigation, especially when layouts change.

• Obstacle detection: The engine uses depth data from sensors (f.e. 3D cameras) to create a real-time map of the surroundings, allowing robots to identify and avoid obstacles.

• Object recognition: This takes robot smarts to a whole new level. Advanced AI algorithms enable robots to not only "see" objects but also understand what they are (boxes, pallets, etc.). This allows them to adapt to changing layouts and handle tasks with greater flexibility.

‍

RGo's Perception Engine explained

‍

Leveraging the NVIDIA ecosystem and powerful computing platform, RGo enables machine learning techniques, such as deep learning-based visual localization. These methods can learn to recognize and adapt to changes in the environment, improving the robot’s ability to localize in dynamic and repetitive conditions. 

‍

‍

‍

RGo's Perception Engine goes beyond standard robot navigation systems.  At its core, it uses a technique called vSLAM (visual Simultaneous Localization and Mapping) to build a map of the environment. But unlike traditional SLAM, RGo's engine incorporates advanced learning algorithms. This allows robots to "learn on the go," adapting to changes in the environment, like new obstacles or rearranged layouts. This makes them significantly more reliable in dynamic factories and warehouses, driving throughput, uptime, safety, and scalability.