Efficient Path Planning for Search and Rescue Missions Using Neural Networks
summary
A Python/Tkinter simulated robot generates a Amalgamated Cost Map (ACM) by merging static map features (road network, buildings, simplified slope) with dynamic evidence (survivor cells = 1, findings scored 0.1–0.5) and trains a UNet neural network to predict ACMs from those layers. The system feeds predicted ACMs into an A-star path planner to route robots toward high-priority cells. Work uses synthetic Kent, Ohio data, is not yet deployed on hardware, and claims to provide a global view missing from standard reinforcement learning.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:50Efficient path planning importance
- 3:36Cost map data types
- 7:56RL challenges in robotics
- 13:21Priority data for unknown regions
- 16:18Amalgamated cost map generation
- 20:14Region visualization
- 23:47Encoding static and dynamic data
- 27:46Neural networks for real-time cost maps
related talks
49mUsing Deep Learning for Perception in Autonomous Systems: A Perspective from Control Theory
32mAI and Minecraft
20mAI on the Fast Lane: Smart Traffic Systems to Unclog our Cities
40mCo-operative robots sharing the load
21mFrom Reactive to Proactive: How AI is Redefining Resilience in Modern Networks
16mOptimized Routing with Satellite Imagery and Computer Vision
49mMonitoring global events with AI-driven satellite swarms
10mChip floorplanning with deep reinforcement learning