In-Network Velocity Control of Industrial Robot Arms
Talkat 19th USENIX Symposium on Networked Systems Design and Implementation (NSDI '22) · 2022-04-04..04-06·
summary
Offloads real-time velocity control into a P4-enabled Tofino switch that stores trajectories as linked-list entries in match-action tables, receives UDP integer/decimal payloads (robot id, joint id, position, velocity), and emits velocity commands at ~2 ms intervals. Prototype uses a six-joint Universal Robots arm, Barefoot/Tofino dataplane and control-plane runtime, approximates control math via longest-prefix matches, handles background traffic with strict-priority queuing, and claims smooth control of 100s of robot arms with convincing precision.
AI-written summary · about
chapters (18)
- 0:00Intro
- 0:29Industry moving to the cloud
- 1:09Setup #1
- 1:21Low level control in the cloud?
- 1:32What about edge cloud?
- 1:52High level control to the cloud
- 3:11Ensuring Real-Time requirements is hard
- 3:37In-network velocity control • Assumptions • Robot arm as a set of actuators
- 4:38Design - Main Components
- 5:36More on the protocol
- 6:54Trajectory representation
- 8:21PID-lik control using approximations
- 9:15Proof of concept implementation
- 10:20Evaluation
- 10:49Precision
- 11:16Effect of background activities
- 12:05Scalability
- 12:27Summary
related talks
18mRobotics: why now?
41mVirtual-Physical Environment Manipulation System based on ROS
13mChallenges in High Performance Robotics Systems
40mCo-operative robots sharing the load
57mThe Convergence of Robotics, the Web, and IoT
14mReliability Engineering for Hybrid Robot-Cloud Systems
40mApplication and Challenges of Using GenAVB/TSN Software Stack for Industrial...
23mMetaArms: Body Remapping Using Feet-Controlled Artificial Arms