Co-operative robots sharing the load
summary
Proposes Autonomous Mobile Programs that migrate whole tasks, make decentralized, selfish move decisions using a simple cost model, and spawn a tiny representative agent to check destinations and cut redundant relocations. Simulation covers 300 machines and ~3,500 AMPs; prototype uses ROS on Raspberry Pi robots (5 then 12). Mobility is weak (task restarts on destination), route-planning tasks take ~15s on one core, Wi‑Fi adds 2–3× latency, and experiments show resilient, emergent rebalancing under robot failures.
AI-written summary · about
chapters (25)
- 0:00Introduction
- 0:37Who am I
- 0:57The problem
- 1:45Rescue robots
- 2:36Large balancing
- 2:52Dont write it
- 4:29Aim
- 6:47Proposal
- 9:05Initial research
- 10:21Representative
- 10:47Load Server
- 11:48Load Agent
- 13:15Analysis
- 17:20What we used
- 17:38ROS
- 18:39ROS statistics
- 19:02How ROS works
- 20:01ROS implementation
- 22:53Fault tolerance
- 24:11Not considering
- 25:07Ramp
- 25:52Experiments
- 27:28Mixed failure
- 28:34Scaling
- 32:45Questions
related talks
20m"Rust in Motion: Building Reliable and Performant Robotics Systems"
36mScaling robots -
27mDeterministic fleet management for autonomous mobile robots using Rust
14mMultiparty Motion Coordination: From Choreographies to Robotics Programs
14mReliability Engineering for Hybrid Robot-Cloud Systems
33mReducing the environmental impact of agriculture with robotics
34mMicroROS, combining advanced robotics and low-cost embedded systems
29mMulti-Agent Pick and Delivery with Capacities: Action Planning vs Path Finding