Follow @LogicalSG for weekday industry reads on robotics, drones & safety, and automation/AI — operator takes, not pitches.
https://t.co/86IUmEZXgc
Drones that hunt drones, with a shotgun on board.
Federal Ammunition, Mossberg and Palladyne AI announced SwarmShot: reusable interceptor drones that split up targets and hunt hostile drones as a swarm.
https://t.co/RNCYixfHA0
Vets, would you trust it? #CounterUAS#Veterans
The hardest thing for a robot to pick up? Lunch. Wet, soft, never the same twice.
Chef Robotics is building on NVIDIA Isaac Sim and cuMotion, and reports up to 60% higher labor productivity at customers.
https://t.co/tENNIpBU96
What's the hardest item on your line to automate?
Is 95% good enough for a robot?
NVIDIA taught robots to assemble GB300 tester trays: 95%+ success on busbars, 90-95% on connectors. NVIDIA says that still falls short of deployment.
At 95%, someone steps in ~1 in 20 tries.
https://t.co/9rXuikurU6
What rate would you accept?
Boston Dynamics handed the keys to an AI leader, not a hardware lifer.
Rohit Prasad, ex-head scientist for Alexa and AGI at Amazon, became CEO Oct 7 to scale physical AI.
The floor test is still uptime.
https://t.co/mZi6sDZpGF
AI or hardware chops: which matters more?
4/ Deals (as reported):
- Multiply Labs: $75M for robotic biomanufacturing https://t.co/GMmfBj1k3f
- Mecka AI: $60M led by Sequoia for robot training data https://t.co/noxCYMqW69
- Teradyne Robotics & Elite Robots settled their dispute (announced Oct 1) https://t.co/WtekhCrCwy
Robots are being put to the test.
Today's Robotics Daily: a welding robot passes its acceptance test, humanoids head to car plants, cobots take spoken commands, and drones eye two more metros.
https://t.co/CwVuOqEDCh
What proof would you need before a robot joins your line?
3/ Walmart plans Wing drone delivery for the Seattle area and Denver in 2027, part of a push toward nearly 20 U.S. metros.
The catch: local approval. County leaders shut down a proposed metro Atlanta hub after 1,000+ residents petitioned (GeekWire).
https://t.co/aKbC6iAwMV
Drone swarms just trained through jamming, at night, in rough terrain.
Swarm Forge at Fort Bragg: 130+ companies cut to ~25. An AI battle manager runs air + ground teams; humans on the loop.
https://t.co/kVTk9ZiZaY
Vets: what would earn your trust?
#Veterans#Military#Drones
A 135 kg robot lost battery power at the top of a staircase and lowered itself to the ground under control.
That's the real test in All3's new Mantis construction robot: how it fails.
https://t.co/owpIPjhTUl
Would you let it on your site?
Will robots really take your job?
Robots take tasks before jobs. WEF projects 170M jobs created and 92M displaced by 2030 across all trends, net +78M.
Someone still installs, programs and fixes the robot.
https://t.co/IDNKPdFQSh
Where would you draw the line?
What one robot failure really costs.
One stalled robot stops the line. Then come missed shipments, overtime and expedited freight.
Siemens: about $2.3M per hour of downtime at a large auto plant.
https://t.co/3WrsEqGmQH
What's the longest one machine has held up your line?
A $69M program is not a production cell yet. The split that matters is onboard AI on a dual-arm versus the cobots that still have to run a normal shift.
https://t.co/nvZCNmdsop
Doosan Robotics just got picked for $69M worth of Korean Physical AI projects.
One target a dual-arm industrial humanoid running AI directly onboard.
The plan also includes 5 kg, 10 kg and 20 kg cobots.
A Korean AI SoC with an NPU will process vision touch force-torque and voice data inside the robot without an external PC or cloud connection.
Before commercial release, Doosan is targeting 72 hours of continuous testing at two or more customer sites, followed by international safety certification.
The second project focuses on nuclear-component welding, targeting ±3 mm path error 95%+ matching with skilled-worker welding paths and over 50% lower working time.
That 72-hour continuous factory test is the detail I’ll be following closely.
Commercial products are targeted for 2031.
@NetChoice A millionth drone package that was not urgent is the telling detail. Once delivery is routine, the cost moves to the ground station and the failed-drop retry, not the aircraft.
A 14 km shore launch into Liverpool Bay is the useful number. The ops question is whether the link budget or the ground-station handoff at the shore was the limiter.
https://t.co/yvpp4UhQog
@neilmcgillivray Upside-down work on steel is a different maintenance problem than a floor cell. The useful number is how many joints a shift can finish, not how high it can climb.
@Newsforce 30 kg of tools on a magnetic climber is a lot of load for a hull joint. I would want the inspection and weld-prep data from one yard section before calling it production.