Robotics In Science Projects

𝗧𝗵𝗲 𝗣𝘀𝘆𝗰𝗵𝗼𝗹𝗼𝗴𝘆 𝗼𝗳 𝗥𝗼𝗯𝗼𝘁𝗶𝗰𝘀 @ 𝗖𝗘𝗦 𝘄𝗶𝘁𝗵 𝗡𝗘𝗨𝗥𝗔 𝗥𝗼𝗯𝗼𝘁𝗶𝗰𝘀! As a trained Tech Psychologist, I often observe Robotics from a very different angle. Not only through: → code → models → hardware → infrastructure …but through human behavior. Because the moment a robot enters a room, our brain instantly evaluates: → Size → Voice → Facial expressions → Eye contact → Movement → Human-likeness → Emotional behavior → Distance & proximity → Predictability And all of this determines one thing: 𝗗𝗼 𝘄𝗲 𝘁𝗿𝘂𝘀𝘁 𝗶𝘁… 𝗼𝗿 𝗱𝗼 𝘄𝗲 𝗳𝗲𝗮𝗿 𝗶𝘁? Some fascinating psychological concepts behind Robotics: 𝗨𝗻𝗰𝗮𝗻𝗻𝘆 𝗩𝗮𝗹𝗹𝗲 The more human a robot becomes, the more emotionally connected we feel to it until it becomes “almost human.” Then discomfort suddenly appears. This concept became one of the most influential theories in Human-Robot Interaction. 𝗔𝗻𝘁𝗵𝗿𝗼𝗽𝗼𝗺𝗼𝗿𝗽𝗵𝗶𝘀𝗺 Humans naturally assign emotions, intentions and personalities to machines even when we know they are not conscious. This is a central research topic in Human-Robot Interaction at places like the MIT Media Lab. 𝗧𝗿𝘂𝘀𝘁 𝗧𝗵𝗲𝗼𝗿𝘆 Studies from institutions like Harvard University show that humans can even begin to overtrust robots especially when machines appear socially intelligent or emotionally aware. 𝗦𝗼𝗰𝗶𝗮𝗹 𝗣𝗿𝗲𝘀𝗲𝗻𝗰𝗲 The more socially a robot behaves, the more emotionally humans respond to it. Which means robotics is no longer only engineering. It’s psychology. And maybe that’s exactly the point many companies still underestimate: People don’t use technology because it is technically perfect. They use it because it feels: → intuitive → safe → emotionally understandable → human-centered If not, even the best technology remains unused. That’s why I believe the future of AI & Robotics belongs not only to engineers… …but also to psychologists, designers, behavioral scientists and Tech Translators. So now I’m curious: Would you trust a robot more if it: → looked human or machine-like? → was taller or smaller than you? → spoke emotionally or neutrally? → had a face or no face at all? How do you feel about robots?

A fascinating breakthrough from Harvard University researchers: soft robots that move using nothing but air pressure. By leveraging advanced 3D printing with rotating nozzles, the team has created structures where motion is “pre-programmed” directly into the material design. Instead of relying on motors or complex electronics, these soft robots change shape and move simply by injecting air. What’s particularly striking is the manufacturing process. Using a combination of flexible materials and sacrificial gels, researchers can print both the outer structure and internal air channels in a single step. Once the gel is removed, what remains is a network of pathways that control how the robot bends and moves. This approach dramatically simplifies production and opens up new possibilities for: Custom medical devices and surgical robots Rehabilitation and assistive technologies Bio-inspired robotics with minimal hardware complexity For those of us working in foresight and emerging technologies, this is another signal of how intelligence is increasingly embedded in materials themselves—not just in software. The future of robotics may be softer, simpler, and far more adaptable than we imagined. #SoftRobotics #3DPrinting #EmergingTech #Innovation #Futures #DeepTech #Robotics #MaterialScience #HealthcareInnovation #Foresight

Very promising! A new open-source platform for research on Human-AI teaming from Duke University uses real-time human physiological and behavioral data such as eye gaze, EEG, ECG, across a wide range of test situations to identify how to improve Human-AI collaboration. Selected insights from the CREW project paper (link in comments): 💡 Comprehensive Design for Collaborative Research. CREW is built to unify multidisciplinary research across machine learning, neuroscience, and cognitive science by offering extensible environments, multimodal feedback, and seamless human-agent interactions. Its modular design allows researchers to quickly modify tasks, integrate diverse AI algorithms, and analyze human behavior through physiological data. 🔄 Real-Time Interaction for Dynamic Decision-Making. CREW’s real-time feedback channels enables researchers to study dynamic decision-making and adaptive AI responses. Unlike traditional offline feedback systems, CREW supports continuous and instantaneous human guidance, crucial for simulating real-world scenarios, and making it easier to study how AI can best align with human intentions in rapidly changing environments. 📊 Benchmarking Across Tasks and Populations. CREW enables large-scale benchmarking of human-guided reinforcement learning (RL) algorithms. By conducting 50 parallel experiments across multiple tasks, researchers could test the scalability of state-of-the-art frameworks like Deep TAMER. This ability to scale the study of the interaction of human cognitive traits with AI training outcomes is a first. 🌟 Cognitive Traits Driving AI Success. The study highlighted key human cognitive traits—spatial reasoning, reflexes, and predictive abilities—as critical factors in enhancing AI performance. Overall, individuals with superior cognitive test scores consistently trained better-performing agents, underscoring the value of understanding and leveraging human strengths in collaborative AI development. Given that Humans + AI should be at the heart of progress, this platform promises to be a massive enabler of better Human-AI collaboration. In particular, it can help in designing human-AI interfaces that apply specific human cognitive capabilities to improve AI learning and adaptability. Love it!

Yesterday, we explored Synthetic Interoception and how robots might gain self-awareness. Today, we shift focus to physical intelligence: how robots can achieve the touch and finesse of human hands. Rigid machines are precise but lack delicacy. Humans, on the other hand, easily manipulate fragile objects, thanks to our bodies' softness and sensitivity. Soft-body Tactile Dexterity Systems integrate soft, flexible materials with advanced tactile sensing, granting robots the ability to: ⭐ Adapt to Object Shapes: Conform to and securely grasp items of diverse forms. ⭐ Handle Fragile Items: Apply appropriate force to prevent damage. ⭐ Perform Complex Manipulations: Execute tasks requiring nuanced movements and adjustments. Robots can achieve a new level of dexterity by emulating the compliance and sensory feedback of human skin and muscles. 🤖 Caregiver: A soft-handed robot supports elderly individuals and handles personal items with gentle precision. 🤖 Harvester: A robot picks ripe tomatoes without bruising them in a greenhouse, using tactile sensing to gauge ripeness. 🤖 Surgical Assistant: In the OR, a robot holds tissues delicately with soft instruments, improving access and reducing trauma. These are some recent relevant research papers on the topic: 📚 Soft Robotic Hand with Tactile Palm-Finger Coordination (Nature Communications, 2025): https://lnkd.in/g_XRnGGa 📚 Bi-Touch: Bimanual Tactile Manipulation (arXiv, 2023): https://lnkd.in/gbJSpSDu 📚 GelSight EndoFlex Hand (arXiv, 2023): https://lnkd.in/g-JTUd2b These are some examples of translating research into real-world applications: 🚀 Figure AI: Their Helix system enables humanoid robots to perform complex tasks using natural language commands and real-time visual processing. https://lnkd.in/gj6_N3MN 🚀 Shadow Robot Company: Developers of the Shadow Dexterous Hand, a robotic hand that mimics the human hand's size and movement, featuring advanced tactile sensing for precise manipulation. https://lnkd.in/gbpmdMG4 🚀 Toyota Research Institute's Punyo: Introduced 'Punyo,' a soft robot with air-filled 'bubbles' providing compliance and tactile sensing, combining traditional robotic precision with soft robotics' adaptability. https://lnkd.in/gyedaK65 The journey toward widespread adoption is progressing: 1–3 years: Implementation in controlled environments like manufacturing and assembly lines, where repetitive tasks are structured. 4–6 years: Expansion into dynamic healthcare and domestic assistance settings requiring advanced adaptability and safety measures. Robots are poised to perform tasks with unprecedented dexterity and sensitivity by integrating soft materials and tactile sensing, bringing us closer to seamless human-robot collaboration. Next up: Cognitive World Modeling for Autonomous Agents.

Why should HR leaders care about video AI breakthroughs? Because it turns out they might be the foundation for a GPT-like breakthrough in robotics. And hence, the automation of manual labor. Veo3 is a model released by Google that is excellent at generating video content. However, there's something much more interesting happening behind the scenes. The model achieves 90%+ accuracy on perception tasks without specific training. Not because it generates prettier videos, but because predicting the next frame forces understanding of how objects move, fall, and interact. It essentially learned how to predict physics! This mirrors what happened with language models. Predicting the next word created understanding. Now predicting the next frame creates >spatial< intelligence. The breakthrough is not about the output quality at all. It's that the model had to learn physics to generate coherent sequences. Now consider the impact on robotics. A humanoid robot using these models doesn't need millions of training runs for each task. It already understands momentum, gravity, object permanence. The Veo 3 paper shows 62 distinct capabilities emerging from video prediction alone. A robot with this model can now pick up a jar, understand which way to twist the lid, and open it - without ever being programmed for that specific jar type. It can throw a ball to another robot and predict where to position its hands to catch it. It can even identify that a hammer should be grasped by the handle, not the head. All from visual understanding alone. We might have our ChatGPT moment for robotics sooner than expected...

Presenting FEELTHEFORCE (FTF): a robot learning system that models human tactile behavior to learn force-sensitive manipulation. Using a tactile glove to measure contact forces and a vision-based model to estimate hand pose, they train a closed-loop policy that continuously predicts the forces needed for manipulation. This policy is re-targeted to a Franka Panda robot with tactile gripper sensors using shared visual and action representa- tions. At execution, a PD controller modulates gripper closure to track predicted forces -enabling precise, force-aware control. This approach grounds robust low- level force control in scalable human supervision, achieving a 77% success rate across 5 force-sensitive manipulation tasks. #research: https://lnkd.in/dXxX7Enw #github: https://lnkd.in/dQVuYTDJ #authors: Ademi Adeniji, Zhuoran (Jolia) Chen, Vincent Liu, Venkatesh Pattabiraman, Raunaq Bhirangi, Pieter Abbeel, Lerrel Pinto, Siddhant Haldar New York University, University of California, Berkeley, NYU Shanghai Controlling fine-grained forces during manipulation remains a core challenge in robotics. While robot policies learned from robot-collected data or simulation show promise, they struggle to generalize across the diverse range of real-world interactions. Learning directly from humans offers a scalable solution, enabling demonstrators to perform skills in their natural embodiment and in everyday environments. However, visual demonstrations alone lack the information needed to infer precise contact forces.

📢 New paper out in #AdvancedMaterials on reprogrammable mechanical metamaterials powered by passive and active magnetic interactions! 🧲 🦾 In this study, we demonstrate how embedding hard-magnetic MREs into architected structures allows for tuning and reconfiguring their mechanical response across static and dynamic regimes. By playing with residual magnetization orientation, stiffness, and external fields, we unlock new pathways toward adaptable, energy-absorbing, and impact-resistant systems. This work opens exciting opportunities in smart structures, soft robotics, and damping systems. Huge thanks to the amazing team and collaborators at Universidad Carlos III de Madrid and Harvard University, and the funding agencies European Research Council (ERC) Ministerio de Ciencia, Innovación y Universidades and monodon! Carlos Pérez García Ramon Zaera Polo Josue Aranda Ruiz Marisa Lopez Donaire Giovanni Bordiga Giada Risso Katia Bertoldi 🔗 https://lnkd.in/dWhanR6t #AdvancedMaterials #Metamaterials #MagnetoMechanics #ImpactEngineering #SmartStructures #ReprogrammableStructures

One of the biggest challenges in robotics and autonomous driving is that actions are not independent decisions happening frame by frame. Real robotic systems operate continuously over time, and every movement depends heavily on temporal consistency. This is one of the reasons why Action Chunking Transformers (ACT) are becoming extremely important in modern robotics research. Instead of predicting a single steering or control output at every timestep, ACT predicts a chunk of future actions together. This allows the robot or autonomous vehicle to generate smoother and more stable trajectories because the model starts reasoning temporally rather than reacting independently at every frame. But there is another important problem. For the same visual scene, there may be multiple valid future actions. A vehicle approaching an intersection may legitimately turn left, turn right, or continue forward depending on intent. A deterministic model struggles here because it tries to collapse multiple valid futures into one average prediction. This is where Conditional Variational Autoencoders (CVAE) become extremely useful. The latent representation learned by the CVAE allows the model to represent multiple possible future behaviors while still remaining conditioned on the observed visual scene. When combined with Action Chunking Transformers, the system becomes capable of generating temporally coherent yet multimodal robotic behavior. Now when language intent is introduced into the pipeline, things become even more interesting. Instead of learning only from vision and action trajectories, the policy can now alter its future behavior based on high-level commands such as LEFT or RIGHT. This moves the system closer toward Vision-Language-Action (VLA) architectures that are now becoming central to modern embodied AI systems. In our latest lecture, we implemented this complete pipeline inside NVIDIA Isaac Lab: -Action Chunking Transformer (ACT) -Conditional Variational Autoencoder (CVAE) -Language Intent Conditioning -Vision-Language-Action pipeline for autonomous driving The implementation was done using Isaac Lab simulation environments along with our TurboPi robotics setup. YouTube Lecture: https://lnkd.in/g3GpRbBP VLA Bootcamp: https://vla.vizuara.ai

📍 From concept to clarity. This is why visualization matters. Here’s a recent case where we modeled a special machine handling bottles into a heated process using robots and a custom conveyor setup. What made it powerful was not just the model itself, but how quickly the customer could understand it. 👀 They didn’t have to imagine how the machine would work They could see it exactly as it would behave in reality That immediate understanding changes conversations, decisions, and outcomes. 💡 The goal of the model was simple Could the system handle the required capacity with the given robot setup? But what we uncovered went beyond that. ➡️ Input spacing was perfectly consistent ➡️ Output spacing varied depending on robot reach and cycle time That variation created gaps. Some small, some large. And that directly impacts downstream processes. ⚠️ These are the kinds of things you don’t want to discover after commissioning. 🎯 With simulation, you can: Identify bottlenecks early, Understand real system behavior, Validate capacity before anything is built Anders Jönsson's team also push this further by introducing variation in the input flow to stress test the system. How much deviation can it handle? Where does it break? And this is where things get really interesting. 🚀 We are moving towards running multiple automated scenarios to explore combinations of variables and system behaviors. Not just ideal conditions, but realistic and even worst-case situations. Because real production is never perfect. 👥 What made this project work especially well was the team A mix of robotic programming expertise, CAD design skills, and a solid understanding of control logic That combination allows us to create: ✔️ Visually strong layouts ✔️ Realistic robot behavior ✔️ Meaningful simulations that actually reflect reality And that’s where the real value comes from. Not just building a model But building understanding before anything is built in the real world Link to the full video here: https://lnkd.in/eTVaXGPw

When researching novel technologies like robotics, we often face a unique methodological challenge: How do we measure scenario realism for a future that does not exist at scale yet? In experimental research, we use written or visual scenarios to manipulate variables that are difficult to test in the real world. For example, my team (Rasoul Mahdavi, Mahsa Talebi, and Halyna Horpynich) ran a study looking at different types of human-robot collaboration. We wanted to understand how people react when working side-by-side with a robot versus when a robot takes on a supervisory role, such as inspecting workflows. Testing these dynamics in advance is critical, but because the concept of a "robot supervisor" feels highly futuristic, traditional realism checks can easily misfire. Usually, a standard realism check simply asks participants if the scenario is "realistic." However, we noticed that participants often interpret this question philosophically. Instead of evaluating the logic of the experiment, they start questioning whether they believe a future where robots supervise humans will actually ever happen. If they disagree with that concept, they give a low realism score, leading the researchers to think that the scenario is ineffective. To solve this, our research team changed how we measure scenario realism. Instead of asking if the situation itself was realistic, we shifted the focus to cognitive immersion and clarity. We started asking whether the scenario was clear, and how easy it was for the participant to mentally place themselves inside that specific situation as if it were happening to them right now. By measuring the ease of imagination and clarity rather than asking if the scenario is "realistic," we achieved much more consistent data. We ensured that participants were reacting to the immediate psychological dynamics of the robot's role, rather than their personal beliefs about the future of technology. Ultimately, our goal is to build the best possible human experiences when integrating these technologies. To get accurate data on how to do that, our realism checks need to measure how well a participant can immerse themselves in the situation right now, rather than whether they agree with the concept itself. #ResearchMethodology #HumanRobotInteraction #ExperimentalDesign #UXResearch #InnovationStrategy