Obstacle and Object Recognition for Mobile Robots using AI

Problem definition

Mobile robots should be able to navigate autonomously and intelligently avoid collision. They should actively recognize the terrain on which they are moving and quickly adapt to it, while avoiding possible obstacles.

Goal

Improving mobility, autonomy and comfort for mobile robots, through the use of AI.

Solution expertise

With the help of trained neural networks, obstacles and objects can be detected in order to avoid collision, while increasing mobility with terrain classification (AI-Frameworks: Mask R-CNN, Detectron2, Regnet-101 and Mobile Net V1). The trained models are further optimized accordingly for deployment on an embedded hardware (AI box). Furthermore, "Follow Me" functionality is enabled through person detection algorithms.

Test bed

Use of the mobile robots and the AI box at the customer's site as a case study (e.g. inspection).