Autonomous mobile robots - move ground platform from point A to B by calculating the optimal path from a map and navigating around obstacles using sensor data
Industry Applications
- Warehouse logistics - Locus, Geek+
- Delivery robots - PostMates, TinyMile
- Lawn mowing - Electric Sheep, Scythe Robotics
- Farming - Monarch Tractor
- Cleaning - iRobot, AvidBots
- Security - Cobalt Robotics
- Self Driving Cars - Cruise, Motional
- Navigation Software - Polymath robotics, MovelAI
Outlook - not so good
- Nuro - layoffs
- FedEx - shutdown
- Amazon Scout - shutdown
- ArgoAI - shutdown
Problems from people I met
- Warehouse robotics CEO - stream large 3D point cloud data to navigate around overhead obstacles (right now they use 2D)
- ML Engineer in Robotic vision company - How to detect lanes from streams
- Tech Lead from delivery robot company - How to develop good distributed systems
Required engineering skills
- C++ - programs need to be fast because they’re real time, 1ms delay can kill somebody like in self-driving cars
- Navigation and Path Planning algorithms
- Distributed systems - sensor nodes publish data to a topic, algorithm nodes calculate best path from that topic, algorithm nodes publish desired move command to another topic
- Continuous Integration and Deployment - robot systems involve a lot of software pieces, and need to be developed in a safe and automated way to not break the full system
Sensors involved
- Radars - high range low accuracy
- Lidars - low range high accuracy
- Computer Vision - obstacle detection, semantic segmentation like image depth