A France-based company named Tornyol is developing palm-sized drones to hunt mosquitoes in the US market, aiming to combat diseases spread by these insects. The drone uses ultrasonic sonar and microphones to detect mosquito wingbeats, rams the target, and patrols five acres per device. The company has received funding from Manifund and Y-Combinator and has raised $500k in seed funding. Despite initial successes, challenges remain regarding battery life, regulatory approval, and scalability. The ultimate goal is to reduce the cost of mosquito control by 100 times to eradicate vector-borne illnesses.
Opportunity charging is gaining traction as robot fleets expand, according to Ben Waters, CEO of WiBotic. WiBotic develops wireless charging solutions and fleet-wide battery management software for industrial mobile platforms. This patented technology allows robots to charge throughout their workflow, improving uptime, return on investment, and safety. The system is applicable to various autonomous systems like AMRs, drones, and humanoids. Ultimately, WiBotic simplifies charging for diverse robots, making it seamless and scalable.
Video Friday features robotics content, including examples of robots collaborating and advanced AI applications. Researchers use video to train physical AI models, such as simulating real-world physics for humanoid robots. Robots are demonstrating complex behaviors, from musical synchronization to navigating rugged terrain in remote locations. Progress is being made in manufacturing complex hardware, like walking robots, through precision assembly. Recent achievements include the Perseverance rover setting a new distance record on Mars.
Robotics is currently in a "Lego era" as its foundational elements like actuators and software are still being rebuilt. This contrasts with the belief that a single general foundation model will trigger a "ChatGPT moment" in robotics. Amulya Vishwanath, an Nvidia insider, leads a venture capital firm investing in physical AI across various fields. She discusses the current state and future investment landscape for robotics. The discussion highlights the difference between current robotics development and the anticipated impact of a general foundation model.
Tau Robotics sends teleoperated humanoids into homes for $30 an hour to collect training data for its AI models. Robots perform vacuuming, mopping, and wiping tasks while remote human operators guide them. The service involves streaming footage to headquarters, which supports the development of the general-purpose robotics model. Physical limitations were observed, as robots struggled with tasks like gripping smooth glass and handling fabrics efficiently. This cleaning service provides valuable, real-world data, though customers have options to request data deletion.
DexForce uses simulated workplaces to train humanoid robots, aiming to reduce the cost of teaching machines useful tasks through synthetic data. The company trains robots in virtual environments, allowing them to handle changes in surroundings, which is argued to be faster and cheaper than collecting human demonstrations. DexForce develops software to simulate physical environments and predict object interactions to generate this training data. The company's CEO believes collecting manual training data is fundamentally inefficient, advocating for simulation to accelerate robot capability. Funding supports this simulation strategy, and their robots are currently deployed in various commercial settings.
1X Technologies' $20,000 NEO humanoid launched globally in 2026, marking it as the first widely available humanoid robot for home use. This robot is designed to manage household chores and boasts autonomous capabilities according to its founder. Despite weighing 66 pounds, it is capable of lifting over 150 pounds and carrying 55 pounds. NEO can autonomously perform tasks such as retrieving items, tidying surfaces, folding laundry, and unloading dishes. The robot's hands are described as dexterous, and it is designed to continuously learn new tasks.