Dr. Judith Müller

🤯🔥 𝘿𝙍𝙊𝙉𝙀𝙎 𝙏𝙃𝘼𝙏 𝙋𝙇𝘼𝙔 𝙋𝙄𝙉𝙂 𝙋𝙊𝙉𝙂? 🏓 Researchers at ETH Zürich have developed an algorithm that allows quadrocopters to hit balls towards predetermined targets. ✅️ Using a Kalman filter, the impact points of the balls are predicted. ✅️ The drag coefficient of the ball and the restitution of the racket are estimated. ✅️ The results are fed back into the algorithm to improve accuracy. ✅️ Demonstrated in experiments with one quadrocopter and two working together. The system has shown improved performance and accuracy in hitting balls to specific targets. ------ 📣Remember to follow me Luis L. if you want more content about #Aviation #eVTOL #Drones #Airports #Innovation or #Tech 👉Follow Negocio Aeroespacial , to learn everything about the aerospace and defense sector. 🔥 Re-share and comment, is the only way for me to have feedback from you 📨Contact me if you need any support, I will love to help you . 👏👏👏 ------ 📲Join our WhatsApp channel: https://lnkd.in/d-wHG2D7 🛰Join our telegram channel : https://lnkd.in/d4GuYeBG

818

66 Kommentare

Modern challenges demand architectures that redefine what's possible. Whether in defense, space, or beyond, the future belongs to those who dare to innovate and disrupt established norms. I was inspired by Uwe Horstmann post on Mithra OS showcasing ARX Robotics’ game-changing solution for upgrading legacy systems in defense. Uwe’s work with Project A Ventures highlights their focus on investing in companies that not only innovate but also bring operational excellence to life. This resonates deeply with the challenges we are tackling at Neutron Star Systems. At NSS, we are redefining mobility in space. With superconductor-based propulsion technology, SUPREME, we are not just bridging the gap between legacy and next-gen systems; we are enabling entirely new possibilities. SUPREME offers scalability, sustainability, and dynamic mobility for future-proof space missions—whether it’s satellite repositioning, orbital debris removal, or cislunar operations. Much like Project A Ventures' hands-on approach to fostering growth in its portfolio companies, we are driven by the need to make space systems adaptable, efficient, and resilient. By integrating advanced technologies like AI and multimode propulsion, NSS is paving the way for a sustainable and secure future in space mobility. This is more than an engineering challenge—it’s about shaping a new paradigm in space operations, one that mirrors the transformative ideas in Uwe’s post. Collaboration across sectors—defense, space, and venture capital—is essential to push boundaries and meet the demands of a fast-changing world. What are your thoughts on bridging the gap between legacy systems and next-gen architectures? Let’s connect and shape the conversation around innovation, sustainability, and absolut resilience. #Innovation #SpaceMobility #Superconductors #DefenseTech #VC #Sustainability #supremetech #deeptech #MSC #deterrence

21

The Munich-based Defence start-up ARX Robotics has introduced an AI tool called ARX Mithra OS, with which already imported military vehicles can be transformed into unmanned ground systems (UGV). According to the manufacturer, the Legacy Autonomy Kit, which runs on the basis of the ARX Mithra OS, allows the retrofitting of a large number of already introduced military vehicles. By digitising parts or the entire fleet, these vehicles could in the future perform numerous tasks without human crew with a high degree of autonomy and a broad-based interconectivity https://lnkd.in/d6zZi97A

3

#quantumcomputing #radar #resourcemanagement Quantum computing addresses the challenge of optimizing radar resource management in dynamic environments, enabling real-time task distribution across multi-sensor networks, which conventional computing cannot achieve effectively.

2

During the extended weekend I took the chance to revisit an old experiment I worked on a few years ago: converting a 360° picture into a virtual reality environment. Contrary to the current hype, not using AI, but pure, hard manual labor 😉 (watch the video for the details). In this case it turned out to work really well for use in a simple R/C flight sim, merging two of my hobbies into one. More than a decade ago I spent most of my evenings working on my own little R/C flight sim in a custom game engine. I even ported it to VR when the Oculus DK 1 came out, but the resolution of the first headsets was just too low for R/C flying: the airplane became a pixel before it reached the end of the runway. I'm happy to say that the current generation is good enough for R/C flying, and in a mobile form factor! The technique I used here is quite labor intensive (although not as much as building an environment from scratch), and has its limitations in freedom of movement. In this case you can easily walk 5 or 10 meters in any direction before the illusion starts to break, but that's also because the picture was taken in the middle of the field. Also, no FPV flying in this environment, but as a line-of-sight simulator, this works pretty well! Do you see other use cases for this technique?

77

9 Kommentare

Delighted to share that our work “A framework for collaborative multi-robot mapping using spectral graph wavelets”, has published in the The International Journal of Robotics Research (#IJRR). This work was led by Lukas Bernreiter in collaboration with Lionel Ott, Roland Siegwart, Marco Hutter, Cesar Cadena and myself. #Opensource Code, Paper and video links are available below. Building consistent maps of large, unknown, and complex environments is a challenging but sought-after capability. The use of multi-robot teams can dramatically reduce the time required to map large scale environments, but on-board pose estimation drift and computational bottlenecks can result in building inconsistent maps between the multi-robot teams and the central base station. For robot teams to efficiently explore and meaningfully collaborate, it is essential that consistent information is readily available between the central base server and the deployed robots, even if there exists mismatch in the information availability and computational resources among the autonomous team members. The central mapping server, receiving information from multiple robots, can leverage its additional compute resources to solve a dense multi-robot pose graph and resolve inter-robot inconsistencies. However, given the on-board computational and bandwidth constraints this may not be a viable solution for each robot to improve its on-board estimation. Furthermore , if a robot is out of communication range of the base station, it may not receive any feedback, degrading its performance during long missions.   In response, we propose a novel collaborative mapping framework that enables global mapping consistency between robots and the mapping server. In our proposed scheme, the central server solves the multi-robot, multi-modal constraint pose graph and broadcasts a computationally-tractable and bandwidth-efficient version of the globally consistent pose graph to the individual robots as feedback. Using graph spectral analysis, at different spatial scales, the robots can detect structural differences between their on-board and the server pose graphs to derive necessary constraints for their fixing their individual robot maps and pose estimates. Our approach specifically identifies the graph nodes corresponding to the origin of the drift, rather than the nodes where the error becomes too large. Furthermore, the updated server graphs can be shared among robots outside the communication range of the base station to enable deep exploration of unknown environments. This work was a collaboration between Robotic Systems Lab and Autonomous Systems Lab ETH Zürich, with field experiments conducted using ANYbotics ANYmal and Voliro robots. Paper: https://lnkd.in/gTeziHXt Video: https://lnkd.in/gW-VgYJ3 Code: https://lnkd.in/gSdhMzUC p.s. the first idea discussion sketch between Lukas Bernreiter and myself from two years ago :D

269

6 Kommentare

IS THERE AN AI / LLM MANIFESTO? I just started co-teaching a class at the BFH Technik und Informatik focussed on giving our engineers hands-on experience with agile. So I thought, instead of the Agile Manifesto, is there something similiar for AI and LLMs? Probably there is. But this was ChatGPT's take on it, which I don't think is half-bad: 1. Human-Centered Design Over Automation for Its Own Sake Prioritize AI systems that enhance human creativity, decision-making, and interaction over purely automating tasks. LLMs should be built to support and complement human work, not to replace it without meaningful improvement. 2. Transparency and Accountability Over Black-Box Models Ensure AI models are explainable and their decision-making processes transparent, rather than deploying powerful but opaque "black-box" systems. This fosters trust and accountability. 3. Ethical Alignment Over Unchecked Innovation Strive for responsible innovation, ensuring that LLMs and AI are aligned with societal values, ethical principles, and legal frameworks, rather than prioritizing speed or scale at the expense of fairness and safety. 4. Continuous Learning and Improvement Over Fixed Solutions Embrace AI models that are capable of adapting, learning, and evolving based on user feedback and changing environments, rather than rigid systems that do not evolve once deployed.

23

5 Kommentare

First day at VCBM2024 in Magdeburg, Germany - some world-class work going on. The image below shows how the team at the University of Magdeburg have developed VR simulation to investigate the effects of different visual and tactile conditions on a navigated needle insertion task. Wow! #VCBM2024 #VisualComputing #VirtualReality

3

As the small aircraft and drones get popular, the enormity of monitoring the lower airspace becomes yet more difficult, the passive radars however are a promising opportunity.

4

Happy to share our latest publication in Acta Astronautica: "Closing the gap between SGP4 and high-precision propagation via differentiable programming" https://lnkd.in/edB7H3Fw. In this work, we present ∂SGP4, a differentiable program version of the widely used SGP4 orbital propagator. By implementing #SGP4 in a differentiable programming framework, we make easier gradient-based applications for a variety of cases, such as orbit determination, similarity transformations, state transition matrix computation, covariance propagation, and also gradient-based optimization to find the #TLE that best matches a given state. All while supporting batch processing of TLEs and modern hardware acceleration (GPUs/TPUs). What’s even more exciting is ML-∂SGP4, a paradigm enabled by the differentiability of our model. Using #ML models within ∂SGP4, users can fine-tune the model using observed or high-fidelity simulation data. As a baseline the model would behave as SGP4, however, if data is available corrections to inputs, outputs and/or parameters can be learned, improving precision while preserving the computational efficiency of SGP4 during inference. The result: an orbital propagator that bridges the gap between traditional models and high-precision numerical methods. The code, with tutorials and usage examples to help you get started is available open-source at https://lnkd.in/e-pa3feS. We hope it will be a valuable resource for satellite operators and researchers. Atılım Güneş Baydin Dario Izzo University of Oxford ESA Advanced Concepts Team #SGP4 #dSGP4 #DifferentiableProgramming #OrbitalPropagation #MachineLearning #AI4Space

116

After my last post about Decision-level code coverage for #Rust I thought it is worth explaining why, unlike our regular tools, our "Zero" tools don't currently support MC/DC-level analysis for #DO178 DAL A / #ISO26262 ASIL D projects. For context, RapiCoverZero uses a branch trace to produce coverage metrics at the object code level. There’s no need for either source code or instrumentation. Hence it can support any compiled language, including Rust. MC/DC - "Modified Condition/Decision Coverage" - is defined in terms of source code level constructs such as conditions and decisions. These don't directly exist in the object code as it only contains branch statements. Object level code coverage can thus only tell you which branches were or were not covered. In order to infer MC/DC coverage from branch coverage you would need a way of mapping branch statements back to conditions/decisions in the source code. This might be possible in some circumstances but is not possible in the general case using off the shelf compilers, which try to remove branch statements where possible as they are bad for performance (branch misprediction can cause an entire pipeline to be invalidated). For example, consider the following source code, which returns the value of bit n in the input value: int check_bit_is_set(int value, int n) { if (value & (0x1 << n))      return 1;    else      return 0; } It contains a decision which means MC/DC is required (even though there is only one condition). Using the GCC compiler for Arm at optimisation level 1 this compiles down to a series of instructions with no branches: check_bit_is_set(int, int):     asrs   r0, r0, r1     and   r0, r0, #1     bx    lr It is thus impossible to infer MC/DC coverage of the source code from branch coverage in this case. More reading:  - About MC/DC -> https://lnkd.in/eRAhcaGB - RapiCoverZero -> https://lnkd.in/e_YqY6ex - Code coverage whitepaper -> https://lnkd.in/eW6Crrg9 #embedded #verification

19

opportunity in Aerospace in Germany

1

You want to learn more about #lasercommunication on #CubeSats? You are interested in the next steps towards satellite based #QKD? Join the Space Challenges Program BOOTCAMP 2024 and let's talk about the advantages of Free Space Optical Communication #FSOC. #OSIRIS4CubeSat #QUBE #CubeISL #Cube1G

46

🚀 ArchiCG Update: Exciting Developments Ahead! Over the past weeks, we've been hard at work improving ArchiCG, our open-source demonstrator for semantic cartography based on interactive compound graphs. Here's what's happening: 🛠️ Current Work: Bug fixes are underway to ensure a smoother experience. We're running a modularization exercise to facilitate extensions and make it easier for developers to contribute. ✨ Looking Ahead – What's Coming in the Next Version? 1️⃣ A view manager on top of a dataset manager to provide consistent data feeds across multiple visual representations, such as: Compound graphs, Timelines, Matrices, and Pivot tables. 2️⃣ Enhanced palettes to support multiple languages, making ArchiCG even more versatile. 3️⃣ Support for polyglot visualizations, leveraging semantic mapping to bridge multiple languages and models seamlessly. 💡 Get Involved! We’re actively preparing ArchiCG for functional and technical evolution and would love to hear from you. Whether you're interested in contributing, extending the tool, or exploring its potential, let’s connect! 📩 Feel free to reach out—your ideas and feedback are invaluable! #OpenSource #SemanticCartography #Interoperability #ArchiCG #CompoundGraphs #Visualization

21

6 Kommentare

Our paper "Mesh-based Object Tracking for Dynamic Semantic 3D Scene Graphs via Ray Tracing" has been accepted for the workshop "Semantics for Robotics: From Environment Understanding and Reasoning to Safe Interaction" at RSS'24. If you are at RSS, meet Lennart Niecksch who will be presenting our work tomorrow, Monday, July 15th. We also prepared a lot of nice video material, Lennart will be happy to show it to you. ---------- Paper: "Mesh-based Object Tracking for Dynamic Semantic 3D Scene Graphs via Ray Tracing" Authors: Lennart Niecksch, Alexander Mock, Felix Igelbrink, Thomas Wiemann, Joachim Hertzberg Abstract: In this paper, we present a novel method for 3D geometric scene graph generation using range sensors and RGB cameras. We initially detect instance-wise keypoints with a YOLOv8s model to compute 6D pose estimates of known objects by solving PnP. We use a ray tracing approach to track a geometric scene graph consisting of mesh models of object instances. In contrast to classical point-to-point matching, this leads to more robust results, especially under occlusions between objects instances. We show that using this hybrid strategy leads to robust self-localization, pre-segmentation of the range sensor data and accurate pose tracking of objects using the same environmental representation. All detected objects are integrated into a semantic scene graph. This scene graph then serves as a front end to a semantic mapping framework to allow spatial reasoning. #RSS2024

15