Thomas Monninger’s Post

Today, we’re starting to share what we’ve been working on: https://hypr.co/ The HYPR team has been heads down building a new approach to robotic intelligence—one centered on a single metric we believe matters most: Learning Velocity. Not how much data you collect. Not how much simulation you use. But how quickly a system turns real-world exposure into intelligence. We believe autonomy doesn’t scale through brute force or brittle infrastructure. It scales by learning faster, closer to reality, and with fewer assumptions. More soon, 🦾 The HYPR Crew — #LearningVelocity #Robotics #Autonomy #AI #OutOfStealth

Today’s robots don’t just follow code -they understand instructions. My final project uses a Large Language Model to interpret the workspace and command a Dobot arm with simple text prompts. From image-based object detection to coordinate transformation, to autonomous motion execution -all driven by AI. RAS 545 has been an inspiring experience, and I’m grateful to Sangram Redkar for his guidance throughout the course, as well as my TAs for their support and mentorship. 🌟 The future belongs to those who can speak both human and machine. #arizonastateuniversity #Roboticsandautonomoussystems #Artificialintelligence

**Introduction to Reinforcement Learning** Ever wondered what powers autonomous driving, game-playing AI, or self-learning robots? I’ve put together a clear and practical introduction to Reinforcement Learning—covering core concepts like agents, rewards, exploration vs exploitation, and key algorithms such as Q-learning, SARSA, Policy Gradients, Actor–Critic, and DQN. The goal was to explain how machines learn by doing, without heavy math—bridging intuition and real-world applications. Link: https://lnkd.in/g8kYNBNY #AI #ReinforcementLearning #ArtificialIntelligence #MachineLearning #DeepLearning #AIAlgorithms #DataScience #AutonomousSystems #Robotics #AppliedAI

*Why Monte Carlo Methods Matter for Agentic AI* Sequential Monte Carlo (SMC) methods, commonly implemented as particle filters (PFs), are a family of algorithms that use random sampling to approximate complex probability distributions over time. They're particularly valuable in agentic AI, where agents must reason about hidden states in partially observable, stochastic environments—such as autonomous robots navigating with noisy sensors or multi-agent systems inferring others' intentions. Unlike exact methods like Kalman filters, which assume linearity and Gaussian noise, PFs handle nonlinear, non-Gaussian dynamics by representing beliefs as weighted "particles" (samples). This particle-based representation makes them well-suited for real-world agentic scenarios. #ParticleFilters #AgenticAI #MachineLearning #Robotics #AutonomousSystems #BayesianInference #ArtificialIntelligence #DataScience #AIResearch #TechInnovation

The robots are coming, and they're getting smarter every day. Not only have they passed the Turing test, but now they're even passing CAPTCHAs. What does this mean for the future of AI and human-computer interaction? It's a brave new world, and we're just beginning to explore the possibilities. #AI #Robotics #Technology #FutureofAI

Revolutionary advancements in AI are happening right here on our campus. Our MIT researchers have enabled robots to "speak objects into existence" with their innovative program, RF-Grasp. This powerful tech uses artificial intelligence and radio waves, enabling autonomous machines to locate, calculate the path, and manipulate the objects they "see." Although challenges unquestionably remain, the potential this technology holds for assisting in automation couldn't be ignored. #AI #Robotics #Innovation #MITResearch Link to the article: https://lnkd.in/ekq49n46

*PRODUCT*: *xAI* showcases *Optimus v2.5* demo, highlighting rapid progress in *robotics* and *embodied AI*. The demo of *Optimus version 2.5* showed improved *motion control*, *task execution* and tighter integration between *AI reasoning* and *physical autonomy*, underlining advances in real-world *robot intelligence*. It reinforces *xAI’s push beyond software*, signals accelerating convergence of *AI models* and *robotics platforms*, and positions *Optimus* as a long-term pillar in next-generation *autonomous systems*. https://lnkd.in/gY9Hx65G #Tech #AI #Robot KaleaVentures

xAI's Optimus v2.5 demo highlights rapid progress in robotics and embodied AI, showing improved motion control, task execution, and tighter integration between AI reasoning and physical autonomy. This signals xAI’s expansion beyond software and positions Optimus as a key pillar in the future of autonomous systems. #Tech #AI #Robot xAI KaleaVentures

Bringing autonomous systems into new environments poses real challenges - especially when data is limited. At The Research and Applied AI Summit (RAAIS) in June, Sasha (Alexandra) Harrison (she/her) from Wayve reflected on how their driving system moved from strong performance in the UK to comparable results in the US after a relatively short period of operation. Her perspective highlights the practical questions facing real-world deployment as autonomy systems expand beyond their original domains. Celebrate the 10th edition of RAAIS on June 12 2026 at raais(dot)co

🤖 Finding hidden objects just got smarter. Researchers have cracked a major robotics challenge that's been tough to solve. Meet Milaps: a new framework that helps mobile robots search environments faster by optimizing their routes in real time. Instead of getting bogged down by complex calculations, this approach uses clever shortcuts to find good solutions quickly—even under tight time constraints. Tests show it outperforms existing methods on large-scale scenarios. • Why it matters: Efficient search is critical for rescue operations, industrial inspection, and autonomous exploration. Better algorithms mean faster response times and smarter robot behavior. Follow for more breakthroughs in robotics and AI! 🚀 #Robotics #AI #PathPlanning #Automation #MobileRobotics #MachineLearning Source: https://lnkd.in/gthPxs6M