Loading post…
NVIDIAElizabeth GoodmanProblem & solution
The Machines that Make the Machines
Summary
NVIDIA’s Seattle Robotics Lab automated GB300 tester tray assembly, focusing on busbar insertion with 16 screws and multi-connector insertion into tight-clearance sockets. The team adopted a “right tool at the right time” strategy, implementing a classical modular pipeline with multi-view FoundationPose for 6D pose estimation, Lissajous-curve manipulation primitives, and high-performance impedance control, later transitioning to DOPER. Work was divided across three robot arms to reduce cycle time, and the system approached the 99.5% success rate and 124-second cycle time targets set by NVIDIA Operations and Foxconn. The key challenges were uncertainty in part geometry and dynamics, and the need for high performance beyond typical research success rates. The project demonstrated that modular systems with robust control can outperform end-to-end learning in constrained manufacturing tasks, challenging assumptions about where learning is most powerful.
Context
NVIDIA’s Seattle Robotics Lab, in collaboration with the Isaac engineering team and with input from NVIDIA Operations and Foxconn, worked on automating the assembly of GB300 tester trays. The project focused on two tasks: busbar assembly and multi-connector insertion, which are currently performed by skilled workers. The team aimed to achieve a 99.5% success rate and cycle times no more than twice that of human workers.
Approach / What changed
The team implemented a classical modular pipeline with separate perception, planning, and control modules. Perception used FoundationPose with multi-view pose estimation, later transitioning to DOPER. Planning used waypoint planners and Lissajous-curve manipulation primitives. Control used high-performance impedance control with automatic damping design. Work was distributed across three robot arms to reduce cycle time.
Takeaways
For constrained manufacturing tasks with uncertainty, classical modular pipelines with robust control can achieve high performance without end-to-end learning.
Multi-view pose estimation using FoundationPose reduced 6D pose errors caused by insufficient object visibility.
High-performance impedance control with automatic damping design enabled gentle yet stable contact during insertion tasks.
FAQ
What success rate and cycle time targets were set for the robotic assembly?
The manufacturers required a 99.5% success rate and a cycle time no more than 124 seconds, which is twice the time of skilled workers.
Why did the team choose a classical pipeline over end-to-end learning?
The classical pipeline was highly effective and straightforward to debug and tune, and no pivot to learning was needed because it met performance targets.
What were the two key challenges in automating GB300 tester tray assembly?
The first challenge was uncertainty in part geometry, appearance, and dynamics; the second was achieving high performance with success rates above 99% and cycle times close to human workers.
Source and freshness
- Source
- developer.nvidia.com
- Published
- Added to Yomu
Summaries are generated from the original by a language model. Check the original before relying on it.
Keep reading
Related reading
The Good Stuff
Paid picks we use ourselves. Links carry a referral code — a subscription through one earns us a commission at no extra cost to you.
The best engineering reads, once a week
One useful email. Unsubscribe anytime.