Enhancing Object Detection & Avoidance Systems for Robots in Challenging Environments Using AI
summary
Researchers adapt off‑the‑shelf sensors and models—Intel RealSense, MK r-c—to improve robot detection in fog, rain and low light, training on synthetic terrain data and achieving 0.7742 accuracy (24/31 correct on a held set). Binary motor classifier uses numerical features (robot velocity, clearance, weight, obstacle distance/height/width) with robot velocity as top predictor. Deployment emphasizes containerized, end‑to‑end automated releases for scalable, repeatable edge inference.
AI-written summary · about
chapters (9)
- 0:00Introduction and overview
- 1:08Challenges in adverse environments
- 2:11Existing solutions and models
- 3:38AI and ML in computer vision
- 4:20AI integration framework for robots
- 5:23Use case and binary classification
- 7:08Container and end-to-end deployment
- 8:31Commercialization opportunities
- 9:56Q&A session
related talks
21mAI, Embedded System and Robotics
21mThe eyes of industry 4.0 - Vision intelligence and machine learning
15mDeterministic, Adaptive, and Sustainable Robotics
01h 02mThe Modern Trolley Problem - Responsible AI Principals
14mApproaches to Object Detection: Edge-Based and AI-Based
36mAI-Powered Real-Time Quality Control
41mKEYNOTE: Image processing, AI, and autonomous systems
26mAI for Autonomous mobile robots