Reinforcement Learning Engineer-Spot Behavior

Boston Dynamics' Spot robot can now accurately read analog thermometers and pressure gauges in industrial facilities using Google DeepMind's new Gemini Robotics-ER 1.6 model, announced on 14 April. The AI model improves instrument reading accuracy from 23% in the previous version to 98%. The model features "agentic vision", combining visual reasoning with code execution to create a visual scratchpad for inspecting images. It also delivers improved multi-view reasoning, allowing robots to use multiple camera streams to better understand their environment. Without agentic vision, the baseline model still achieves 86% accuracy. Google describes Gemini Robotics-ER 1.6 as its safest robotics model yet, with improved capacity to follow physical safety constraints and perceive injury risks to humans.

Why robots at a phone conference? Because 6G changes everything. 6G isn't just about faster downloads. It's about turning robots into intelligent fleets that think, learn, and coordinate in real time. I walked the halls of Mobile World Congress last month expecting the usual suspects: new phone announcements, network upgrades, maybe some predictable demos. What I didn't expect was humanoid robots everywhere. The question kept nagging at me as I moved from booth to booth: why are there so many robots at a conference obsessed with mobile phones? Sure, a dancing robot draws a crowd. It's good theatre. But to dismiss them as mere publicity stunts would be missing something genuinely significant happening underneath the spectacle. There's a real conversation unfolding about what happens when robotics meets next-generation connectivity, and it's worth paying attention to. The missing link between two worlds. On the surface, 6G and robotics seem like unrelated futures. One is a network. One is a machine. But that's where things get interesting. According to experts at the conference, 6G won't just make your phone download speeds faster. It'll transform robots from standalone mechanical tools into coordinated fleets that share information, learn from each other, and operate as part of a larger ecosystem. This isn't speculation about something distant either. Companies like Boston Dynamics and Honor have already shown off humanoid robots designed for real work in factories and homes. But there's another level to unlock, and it requires the connectivity that 6G promises to deliver around 2030. The shift sounds subtle in theory. In practice, it's the difference between a robot that can only do what it was programmed to do and a robot that can adapt, learn, and collaborate in real time. How 6G becomes a robot's nervous system. Let's break down what makes 6G special for robotics. First, it acts as a sensor network. Qualcomm's executive vice president of Robotics, Nakul Duggal, explained to attendees how sensors embedded in robots and environments would allow the 6G radio to function like radar, constantly scanning and mapping surroundings in real time. A crowded room stops being a navigation nightmare when you have that kind of environmental awareness. But sensing is just part of it. The real power comes from pure speed. Today's 5G networks aren't designed to handle the continuous AI requests that smart robots demand. Frank Long, associate director of intelligent services at Cambridge Consultants, noted that "with 6G you can pretty much have that quality of service guarantee" that robotics needs. The demo his team brought showed a humanoid robot that could pick up and place a box based on where you pointed, adjusting its grip in real time as it reacted. That required enormous compute power, and it only worked because they used a private 5G network. Public networks today can't reliably handle that workload. This matters because robots won't be working alone. They'll need to communicate constantly with infrastructure and with each other. Imagine a retail scenario that Anshuman Saxena, general manager of robotics at Qualcomm, described: one robot unloading soda cans from a truck while another restocks shelves. For them to work efficiently, they need to share what they see. The unloading robot needs to tell the shelving robot how many cans are coming. The shelving robot needs to know what space is available. This kind of coordination, sometimes called long-horizon planning, is where 6G becomes essential. The home scenario you're not thinking about yet. Here's where it gets weird in a good way. You might imagine a single humanoid robot in your home as being totally different from multiple robots in a warehouse or store. It's not, really. Your phone already coordinates with your security cameras, your smart lights, and your other devices. A home robot would just be another device in that mesh. Or maybe you'll have one humanoid and several smaller robots designed for specific tasks. The point is that even in your private space, there's a "fleet aspect" happening. You don't feel it, but it's how modern devices work. What's more interesting is the learning part. Your phone collects data about how you use it, and that data gets fed back to make the software better. A 6G-equipped robot would work the same way. It might learn to serve coffee in a hotel for months, then come to your home with that training already embedded. But it'll still need to learn your specific layout, your quirks, your preferences. The robot that serves coffee in a thousand hotels will bring all that collective knowledge to your kitchen. The heat problem and why it matters. Here's a detail that stuck with me. A robot might serve you a cup of coffee without understanding that the cup is scalding hot. Humans react instantly to heat. Infeeds pull its hand back without thinking. A robot needs to be taught this through data. Lots of data. From real situations. Saxena pointed this out as an example of why continuous learning matters so much. Right now, gathering that kind of real-world training data and feeding it back to improve the system is incredibly slow. Networks bottle up. The process becomes expensive and time-consuming. 6G removes that bottleneck. The speed and reliability it offers means robots can actually gather and process the information they need to become genuinely safe and useful in human environments. But here's the honest part: this is going to be hard. Frank Long put it perfectly: "Put it this way, members of my immediate family still struggle with opening the baby gate in my stairs, even after extensive training. So a robot, I think, might be a few years away from opening that baby gate." It's funny, but it's also a reality check. Even with 6G on the horizon, there's a lot of work to do. What about right now? Companies aren't waiting around for 6G to show up in 2030 and beyond. Qualcomm is working with robotics firms like Neura Robotics, pushing the boundaries of what current Technology can do. The robots being deployed now are learning, improving, and getting better with dexterity and problem-solving. They're priming themselves to take full advantage of better connectivity when it arrives. The convergence of robotics and 6G feels inevitable now. But there's also something worth noting in all this: the gap between what's possible in a controlled demo and what's actually safe and useful in a home is still enormous. The robots coming to warehouses and hotels in the next few years will be genuinely useful. The humanoids heading to homes? That story might be best written in the 2030s, if not later. Which raises a question worth sitting with: are Infeeds ready for that kind of automated assistance in its most private spaces, or are Infeeds just moving fast because Infeeds can?