Scientific Systems的动态

Scientific Systems Conducts Groundbreaking Research Pairing Cutting-Edge Autonomy with Search and Rescue Dogs COSMIC-T System Analyzes Canine Behavior to Locate Victims Faster During Live Search and Rescue Training Missions WOBURN, Mass., Sept. 24, 2024 /PRNewswire/ -- In a training exercise conducted for the Defense Advanced Research Projects Agency (DARPA) utilizing federal canine search and rescue teams, Scientific Systems demonstrated groundbreaking artificial intelligence software that enables robotic teammates to understand the animal's search response behaviors and accurately predict the location of a lost victim. This first-of-its-kind demonstration represents a breakthrough in teaming capabilities pairing autonomous systems with living teammates. Utilizing an uncrewed aerial vehicle, the company's COSMIC-T (Collaborative Intelligence for Olfactory Search Missions Integrating Canines and Technology) behavior analysis algorithms fused data from each canine's GPS collar with the environment's topographical and meteorological information to generate and refine real-time victim location predictions. Those predictions can be conveyed to other members of the search team or used to direct the searches of autonomous UAV teammates to locate victims faster. "We are extremely encouraged by the results of COSMIC-T's collaboration with the federal search and rescue dogs during these live training exercises," said Ssu-Hsin Yu, Vice President of Technology Innovation at Scientific Systems. "By teaming autonomous systems with humans and non-authoritative teammates like dogs, we greatly improved the effectiveness of search and rescue operations. This breakthrough demonstration of AI and real-time data-fusion demonstrates COSMIC-T's ability to improve outcomes with collaboration between living partners and technology." COSMIC-T's canine training data was captured during training missions conducted by the Boone County Missouri Fire District, the sponsoring agency of the Federal Emergency Management Agency (FEMA) Missouri Task Force One (MO-TF1). "It's been a privilege to assist Scientific Systems in its COSMIC-T program," said Scott Olsen, Fire Chief, Boone County Fire Protection District. "Canine Search Teams are one of the most important components of our disaster response capabilities, and we're proud to have contributed data and expertise towards technologies that can help advance the Search and Rescue mission." COSMIC-T builds upon the success of Scientific Systems' DARPA I2O Teammate Aware Autonomy (TAA) program, which in 2020 demonstrated the world's first successful execution of a search and rescue exercise using a trained search and rescue canine and Scientific Systems' autonomous RAPTOR UAV without any instruction or intervention by a human handler. Link: https://lnkd.in/eh7ViYU6

"Ethical Decision Making in Contested Areas Elevated by OSNet Enhanced Special Operations UAVs" David Cannon 2024 Imagine AI software integrated into unmanned aerial vehicles (UAVs) that not only identifies an individual in contested areas but also informs you of who they are. This is the groundbreaking capability offered by the integration of OSNet, an advanced neural network architecture, into UAV systems, redefining the landscape of special operations and tactics. OSNet's omni-scale feature learning excels in facial recognition, providing unparalleled precision in identifying both combatants and hostages under diverse and challenging conditions. This AI technology, when deployed via UAVs equipped with cutting-edge sensors and supported by encrypted communications, ensures the seamless, real-time exchange of critical data. Such integration empowers special operations forces with actionable intelligence, enabling precise identification and understanding of individuals' roles within a contested domain. This advancement significantly enhances tactical flexibility and operational safety for special operations teams. When layered with a multi capability AI package, It supports the execution of precision strikes and targeted rescue missions with minimized risk to civilians and maximizes the safety of non-combatants. The real-time processing and comprehensive operational awareness facilitated by this technology allow for dynamic tactical adjustments, crucial for the success of missions in high-stakes environments. Addressing the challenges of hardware compatibility, data security, and ethical deployment must be essential considerations when developing defensive applications. As UAVs with OSNet AI become integral to modern warfare and peacekeeping efforts, ensuring their kenitic use adheres to strict ethical standards must also be paramount. The future of special operations, augmented by UAV-integrated OSNet technology, promises a new era of mission efficiency and effectiveness. This AI software does not just identify individuals—it provides a deep understanding of who they are, offering a tactical edge in both combatant identification and hostage rescue scenarios. This capability sets a new benchmark for conducting operations with precision, safety, and ethical responsibility in contested areas.

Exciting news! Team RAINDROPS is selected for the XPRIZE Wildfire competition with the goal to develop a novel system-of-systems based on aerial robots to autonomously detect *and* suppress incipient stage wildfires at rapid response times! Key is the goal to suppress the wildfire, not "merely" detect and localize it. Who we are:?RAINDROPS is a student-driven collaboration between the Norwegian University of Science and Technology (NTNU, Norway) and Brigham Young University (BYU, USA), with the option of including new partners. RAINDROPS seeks to investigate the potential of utilizing a team of uncrewed aerial systems (UAS) that features the low costs and scalability necessary to facilitate incipient-stage wildfire detection and suppression over large areas with rapid response times. Our mission:?The envisioned solution combines two classes of systems, namely those tasked with observation and fire detection (“Observation UAS”) and those tasked with fire suppression (“Fire-suppression UAS”). The first set of robots emphasizes long-endurance and exploits a multi-modal sensing payload, sensor fusion algorithms, and deep learning to detect and localize incipient-stage wildfires rapidly. The platforms will implement meteorologically-informed path planning and multi-agent teaming such that the fire search and detection process is autonomous. The UAS tasked with fire suppression integrate a specialized payload that is released and dropped on an identified fire location to suppress it. With the location of an incipient-stage wildfire detected by the Observation UAS and communicated to the team, Fire-suppression UAS are tasked to rapidly fly towards the area of interest to deliver their suppression payloads. The complete system-of-systems emphasizes autonomous operation and scalability of costs to facilitate adoption. Learn more about the selected teams: https://lnkd.in/guGnqRds Follow our updates through our website: https://lnkd.in/gmBMpgEm Tor Arne Johansen Randy Beard Tim McLain #robotics #autonomy #wildfire #xprize

Looks interesting.... some kind of autonomous tracking and terminal guidance. Def Stealth Industries’s Next Gen Real Terrain Multi Spectral Camouflage solutions need of modern battle field.

This week in Human Factors in Combat, we covered Remote Killing. One of the questions we discussed this week was whether these disruptive technologies: drones, cyber, GenAI, have significant changes on warfare or is this just the latest shift in how war is waged that has been ongoing for millennia. There certainly are arguments for both sides. We were fortunate to have CAPT Brian O'Lavin, one of our senior NSW reps on the Yard and Mr. Mauro Mujica-Parodi III (former USMC infantry officer turned tech executive) teaching this week's curriculum. There was quite a bit of material to discuss with this topic: AI-powered targeting (pro/cons of the human in the loop), autonomous weapon systems, swarming drones, and cyber-enabled remote killing. We talked about the key technological advancements that have driven the rise of drones, AI, and autonomous systems. We looked at a number of videos from Ukraine and what is happening there and where things could go with China. We talked about the rise of cyber-warfare and the remote non-physical killing that is ongoing. In regards to this topic, I don't think this is terribly controversial, but my personal opinion that I shared with the students is that we are at war already with these matters. It's a war that is currently non-physical (doesn't involve boots on the ground in the classical sense) but it certainly is being waged globally with our near-peer competitors as well as non-state actors. Every now and then, there are traces of this that make the news but much of it is currently waged under public awareness. This is the world they will be graduating into. The experiential activity this week was challenging and had deep learning for the students. I had all 50 students in the class meet at the Naval Academy's endurance course, which is a 2+ mile path with about 15 obstacles along the the way in the woods of Naval Station Annapolis. The students were broken into 5 teams and had to run the endurance course while the Naval Academy's SWAT-C team (Squad With Automated Teammates Challenge)?attacked the students with Skydio Drones (thank you Adam Bry!) The SWAT-C team utilized natural open areas and choke points along the course to target the teams. If they were able to hover directly above a student and capture them within the frame of a picture, this simulated a munition drop, and an instructor would mark them as "killed." The teams had to compete against each other in completing the course with the fastest time while keeping their members "alive" pitting the balance of speed versus security. My group took heavy casualties with 7 of our 8 members taken out by drones over 45 minutes on the course. What was remarkable was watching the team quickly develop their tactics on the fly to get through the course. Some pics and video from the event can be found here: https://lnkd.in/ed2fd-Ju

Looks interesting.... some kind of autonomous tracking and terminal guidance. "AuterionOS application that allows minimally trained operators to designate static and moving objects in video, and enter an automatic flight mode that intercepts tracked objects with low CEP, even if link is lost during the #terminal phase".

Special Focus on UAV Swarm Autonomous Control Unmanned aerial vehicle (UAV) swarms have attracted significant attention in both military and civilian applications, including battlefield monitoring, traffic control, border patrol, disaster rescue, agricultural monitoring, and target tracking. The autonomous control of UAV swarms governs their collective behaviors, which in turn affects their task performance. This special focus highlights new approaches in UAV swarm autonomous control theory, techniques, and systems. (Link: https://lnkd.in/gvShJBG4) It includes five papers covering the following topics: ? (1) Prof. Hugh Hong-tao LIU et al. present a path planning algorithm for monitoring crop fields using fixed-wing unmanned aerial systems. Case studies show that the proposed algorithm effectively plans routes for convex-polygon-shaped crop fields. ? (2) Prof. Andrey POLYAKOV et al. focus on the optimal rejection of bounded perturbations in a linear leader-following consensus protocol. Simulation results demonstrate that the multi-agent system (MAS) state can converge to a minimal invariant ellipsoid even in the presence of bounded disturbances. ? (3) Prof. Yunbo ZHAO et al. present an autonomous multi-drone racing method based on deep reinforcement learning. Extensive comparisons with state-of-the-art methods in simulations and real-world experiments confirm the effectiveness and superiority of the proposed method. ? (4) Prof. Haibin DUAN et al. propose a UAV swarm air combat maneuver decision-making method based on multi-agent reinforcement learning and transfer learning. Simulation testing and ablation experiments demonstrate the effectiveness of the proposed method in dealing with swarm air combat scenarios. ? (5) Prof. Mou CHEN et al. focus on dynamic event-triggered fault-tolerant cooperative resilient tracking control with prescribed performance for UAVs. The proposed control scheme demonstrates viability and efficacy through numerical simulations. ? Guest Editors: Haibin Duan, Beihang University Hugh Hong-tao LIU, University of Toronto

Auterion is proud to publicly announce a complete solution for automated terminal engagement with attritable unmanned systems. Our most fundamental goal is to make autonomous computing ubiquitous, flexible, and interconnected for the industrial and defensive benefit of liberal democracies. The Auterion hardware and software platform has been engineered to support every unmanned system and every onboard application that serves this mission. We have worked with government and OEM partners in the US, EU, and Ukraine over the last several months to develop this new capability, which has now reached TRL-9 following hundreds of test flights, as well as successful front-line engagements by our Ukrainian partners. ???? While the guidance math was hard, the moral calculus was not. Auterion T&I (Track and Intercept) is an AuterionOS application that allows minimally trained operators of low-cost UAS to designate static and moving objects in video, and enter an automatic flight mode that intercepts tracked objects with a CEP rivaling loitering munition systems costing >$100k. ?? GPS is not used. Unlike traditional FPVs, the drone will continue into the target after the datalink is jammed. The tracking and guidance is sufficiently precise to begin engagements at significant distances to target, outside the range of jammers. ?? This capability is not just for FPV type quadcopters. We have engineered the tracking and terminal guidance applications to also work with larger fixed-wing platforms from our partners. While Auterion T&I is compatible with the existing Skynode, today we are also announcing Skynode S. Skynode S is specifically designed to enable deployment of unmanned systems running AI and computer vision applications at quantities that can make a strategic level impact. ??Skynode S is a compact, low cost, chip-down design that includes both an FMUv6x flight controller and a small but powerful mission computer with a dedicated Neural Processing Unit for AI and computer vision applications. ??Skynode S is NDAA compliant and built for integration, featuring standard 30.5mm x 30.5mm 4-hole mounting and direct 8S LiPo/LiIon input without a power module. Plus it has all the right I/O, including MIPI-CSI and USB 2.0/3.0. Finally, Auterion has developed a reference design kit for UAS, called Dragon. ??The Dragon Reference Design Kit helps drone manufacturers be up and running quickly with systems that support up to 2.5 kg payload capacity, with the ability to complete missions up to 25 km away with a 1.5 kg payload. Dragon is already making a difference on the battlefield, and we were proud to offer it as the reference design kit for the recent hackathon in Krakow. Read the full press release: https://hubs.ly/Q02DC97T0 Learn more about the products here: https://hubs.ly/Q02DCd2Y0

VERTLINER is a field #robotics #startup company specializing in the #development of #autonomous unmanned #aerialsystems, for #inspection and #monitoring activities, mainly in the #construction #industry. "The field of UAS #development presents an extensive array of practical #applications across various industries. Asset inspection, underground mapping and exploration merely scratch the surface of its potential. Given the adaptability of UAS customization, virtually any challenging or hazardous task stands to benefit from increased #safety and #efficiency, ultimately conserving valuable resources and, increasing the safety factor for humans in a synergistic manner.?" - Konstantinos Gonidakis, PR & Communications Manager at #Vertliner. https://lnkd.in/dJJk6tw4

Excited to share the findings of our recent study on automating sheep counting using unmanned aerial vehicles (UAVs) and machine learning techniques. Traditional counting methods are labor-intensive and disruptive, while current automated solutions struggle with detection accuracy due to low object-to-image-pixel ratios and high object densities. Our research introduces a novel sub-window inference technique, which significantly improves detection and counting accuracy by increasing the object-to-image-pixel ratios. The method sub-divides the images into smaller windows on which inference is done, and the results are aggregated together in a way that rejects duplicate detections. Compared to four other methods, our approach achieved a mean absolute error (MAE) of 3.21 sheep and a mean absolute percentage error (MAPE) of 1.27%. We also found that incorporating random cropping data augmentation during model training further enhances accuracy. While sub-window inference does not facilitate real-time predictions, it substantially reduces the MAE in automated sheep counting, providing a highly effective solution for sheep farmers. Well done to David Biggs for the great work, and Kristiaan Schreve for the co-supervision. #UAV #MachineLearning #AgricultureTech #SheepFarming #Innovation https://lnkd.in/dvkmwBHx