Logical Solutions Group Robotics Daily — Monday, October 19, 2026
Five private-sector robotics stories published October 18–19, 2026: a cobot maker puts a self-driving veteran in charge of its robot software, AI vision that picks jumbled food packs at PACK EXPO, a physical AI deal aimed at Japanese factory lines, a digital twin that coordinates mixed robot fleets, and early money for factory robot hands. It was a light Monday after the weekend, so we ran five stories rather than repeat older news.
1. Doosan Robotics names a former Hyundai self-driving chief as CTO and dates new robot software products for 2027
MEKAWIRE (Doosan Robotics release) | October 19, 2026
Doosan Robotics said Jang Woong-jun, former head of Hyundai Motor's autonomous driving center and ex-chief strategy officer of Motional, is its new chief technology officer, and sales head Kim Hyun-soo is chief strategy officer. Jang will lead Agentic Robot OS, software meant to let a robot perceive its work area, optimize its paths, and carry out tasks without step-by-step programming. He will also deepen the company's NVIDIA partnership and bring self-driving perception methods to its industrial humanoid. Products using the software are planned for 2027, and the humanoid for 2028. MEKAWIRE's review of filings shows first-half 2026 revenue of KRW 32.97 billion, an operating loss of KRW 26.49 billion, and North American solutions supplying 53% of revenue.
Why it matters: Cobot arms keep getting cheaper, so makers are betting that margin moves to software that cuts setup time. Doosan has now put dates on that bet. Before planning around it, ask for a specification, a price, and a reference customer; none has been published yet.
New to the field? A robot operating system is the software layer that handles sensing, planning, and motion, so each new job does not have to be programmed from scratch.
Source: mekawire.com/news/doosa…
2. JLS Automation debuts AI vision that lets its packaging robots pick touching and stacked food items
Robotics 24/7 | October 18, 2026
At PACK EXPO in Chicago (October 18–21), food-packaging robot maker JLS Automation is giving JLS VISION AI its live debut after first previewing it in September. The system detects and classifies products, works out which item is on top and pickable, and keeps reacquiring targets as products shift, so a JLS robot can turn touching, overlapping, or stacked items into an organized flow. It processes images at up to 15 frames per second, supports 2D and 3D vision, and targets variable products such as raw proteins, frozen foods, baked goods, and flow-wrapped packs. It works only with select JLS robotic systems and is not a retrofit for other equipment. The show demo picks condiment packets from irregular flow.
Why it matters: Many food lines pay for extra sorting equipment or extra people just to space product out for a robot. Vision that tolerates messy presentation attacks that cost directly. Ask for pick and miss rates on your own products, not the demo item.
New to the field? Pile picking means a robot choosing and lifting items from a jumbled heap instead of from neatly spaced rows.
Source: robotics247.com/article…
3. Tokyo Boeki signs an MOU to bring ROAI's robot-planning AI to Japanese factories
PR Times (Tokyo Boeki Holdings) | October 19, 2026
Japanese trading group Tokyo Boeki Holdings signed a memorandum of understanding with Seoul-based ROAI, which uses spatial-intelligence physical AI to automate process design for industrial robots. According to the release, ROAI's technology has cut cycle times and engineering hours on welding and inspection lines at automakers including Hyundai Motor, Kia, and GM Korea, and ROAI has raised more than $10 million in a Series A round. Tokyo Boeki will find customers and plan proofs of concept in Japan; ROAI will supply the core software, evaluation licenses, and technical support.
Why it matters: Engineering and commissioning often cost more than the robot itself, and tools that generate robot programs automatically go straight at that cost. An MOU is not an order, and the savings were not quantified. Watch for paid pilots with before-and-after engineering hours.
New to the field? Robot process design means deciding where robots go, what path each one takes, and in what order they work. Specialist engineers usually do this by hand.
Source: prtimes.jp/main/html/rd…
4. Reply earns NVIDIA's physical AI specialization, pointing to a digital twin that runs mixed robot fleets for Otto Group
Reply (company news) | October 19, 2026
IT services firm Reply says it is among the first NVIDIA partners to earn the Physical AI Specialization, awarded after a technical review of a customer project, its solution architecture, OpenUSD skills, and roadmap. The featured project is at Otto Group's Hermes Fulfilment center in Loehne, Germany. There, Roboverse Reply built a high-fidelity digital model that underpins Otto's Robotic Coordination Layer, which connects autonomous mobile robots from different makers with fleet-management systems and the central warehouse management system. New layouts, robots, and workflows are simulated in NVIDIA Omniverse and Isaac before they go live.
Why it matters: Big warehouses rarely run one robot brand, and the hard part is making mixed fleets share floor space and work orders. Testing changes in simulation first lowers the risk of adding a new fleet. The Otto project was disclosed earlier; the certification is today's news.
New to the field? A digital twin is a live virtual copy of a facility, used to test changes before making them in the real building.
Source: reply.com/en/newsroom/n…
5. Gewu Tongzhi raises nearly RMB 100 million to build AI for dexterous factory robot hands
TechNode (citing IT Home) | October 19, 2026
Hangzhou-based Gewu Tongzhi, founded in 2026, closed an angel round of nearly RMB 100 million, according to reports published October 18. Investors include Hongfu Asset Management, Peak Investment, Hangzhou Capital, and Hangzhou Talent Group. The company builds AI for dexterous robotic manipulation, combining world models with learning from physical interaction. Founder and CEO Cai Jia previously worked at Huawei's research labs and automotive business on its ADS driving systems. The money will go to model training, hardware, hiring, and real-world data collection. Gewu Tongzhi is running proofs of concept with customers in textiles, optical modules, and automotive components, with a first delivery expected by the end of 2026.
Why it matters: Investors are still funding the robot brain for fiddly handling jobs that fixed automation struggles with. This is very early: the real test is the first paid delivery and how it performs on a live line.
New to the field? Dexterous manipulation means handling objects with finger-level control, such as inserting, threading, or folding, rather than simple pick-and-place.
Source: technode.com/2026/10/19…
What to watch
PACK EXPO in Chicago, through October 21: measured pick and miss rates from AI-vision and palletizing demos, not just show-floor footage.
Doosan Robotics' third-quarter results in November, and any first technical disclosure or named customer for Agentic Robot OS.
Whether Tokyo Boeki and ROAI name a first Japanese factory for a proof of concept, with before-and-after engineering hours.
Sources are linked above. Figures are as reported by the cited publications and companies and have not been independently verified.