KingKongRobotics

Jumper: the open-source crab robot Jumper is an open-source robotic crab from KingKong Technology. Its compact servos use two encoders to track the motor and joint. Footage: KingKong Technology — kingkong.tech Simulation: KingKong Robotics, modified from docs/media/dance.gif — KingKongRobotics/jumper (Apache License 2.0; see repository LICENSE and NOTICE). #Robotics #OpenSource #Technology

Jumper, the open-source crab robot : In Search of Sun [V2] Jumper likes to see things from different angles. As a result, he sometimes finds himself in rather tricky situations. But every problem is an opportunity to learn - for him, and for us! Why this [V2]? Because some transitions could have been smoother, some scenes still needed a bit of fine-tuning in terms of timing, and a few details deserved to be more visible. And also because my son and I - being a little sadistic - hadn't yet found the time to try something on Jumper that had been on our minds for quite a while. Well, now we have! We have to admit, Jumper is an awfully good sport. 🦀✨ Did you spot all the details on Jumper's displays? Let us know in the comments! Many thanks to @KingKongRobotics and Jumper’s creators for their brilliant robot and the resources they have shared. We hope you enjoy watching it as much as we enjoyed making it! ---- 🛠️ What we added to the original model, or tested (based on our initial observations of its specs): • Solar panels, • Sunny spot detector (technical term: "prime nap zones"), • LED strips and their animations, • Back-walking (just in case), • First-person view and front display repurposing, • Unprecedented yet essential robotic features, with console status logs: purring, snoring, stretching, and paw-count checks, • AI features designed for Jumper: object detection powered by YOLO and a budget-conscious reasoning mode. When we first heard about the Jumper Crab Challenge, my son and I just looked at each other: "We're doing this!" This is my first attempt at creating a 3D animated short film, based on a story created with my son. From the storyboard, sets and Jumper’s appearance to animation, editing and sound - including the little details scattered throughout the scene and across the screens - the project was a process of learning, experimentation and refinement. Codex assisted with programming and automation, while I guided the choices that shaped the film. Optimizing hardware resources, developing ideas, giving detailed instructions, reviewing results, troubleshooting issues, and deciding what to keep or change required substantial human work and many iterations. Bringing the story to life meant paying close attention to expression, continuity, pacing, and the smallest details. All robot simulations, 3D scene creation, rendering, video editing and sound processing were carried out locally. The workflow used Python scripts developed with Codex assistance, MuJoCo for robot simulation, Blender for 3D scenes and rendering, and FFmpeg for assembly and encoding. We started with the official Jumper model and existing locomotion and posture tools. Walking and expressive gestures were simulated in MuJoCo, checking joint limits and ground contacts before transferring the resulting movements to Blender. Some imagined features and technical additions reflect what seemed technically plausible to me, rather than confirmed capabilities. The screens were also used creatively, without knowing their actual technical specifications. Initially, only the NVIDIA GPU rendered the 3D images. We brought the integrated AMD GPU into use, improved CPU task allocation, and optimized concurrent rendering on NVIDIA. We estimate an overall rendering throughput increase of approximately 39%, with unchanged quality settings. These experiments highlighted the importance of balancing the workload across available resources. ---- CREDITS Concept, story, Art direction: TonioBriss AI-assisted production with OpenAI Codex, under human direction and review. Tools: Blender/Cycles, MuJoCo, Python, FFmpeg, NumPy, SciPy, Pillow, Audaspace. Jumper: Copyright 2026 KingKong Robotics. Official robot geometry and display expressions adapted under the Apache License 2.0. Changes include black-and-purple styling, solar-panel and LED treatments, screen graphics and story-specific animation. • KingKongRobotics/jumper • apache.org Visual resources: • “Velour Velvet”: colormass & Rico Cilliers / Poly Haven (CC0) polyhaven.com • Countryside photograph: Gary Houston / Wikimedia Commons (CC0) commons.wikimedia.org • Fonts: DejaVu Sans & DejaVu Sans Mono • Object-detection graphics inspired by YOLO (You Only Look Once). Sound and ambience: • CC0 recordings: Joseph SARDIN / LaSonotheque, Kenney, Glitchedtones, Flem0527, deleted_user_2906614, mista bojjob, fvcalderan, odilonmarcenaro, kyles, ngruber, Burningmonkey, Joao_Janz, KoimeMatsuchiya, mickdow, and Saaabaaato. • Public domain recordings: Oona Räisänen (Mysid) – blackbird song; Adam Cuerden – purring. • Sources: LaSonotheque, Kenney, Freesound, OpenGameArt, Wikimedia Commons. Additional scenery, graphics and sounds created for this film. No music. Thank you to Jumper’s creators and the contributors behind these tools and resources! #Jumper #CrabRobot #OpenSourceRobot #Blender3D #robotics #mujoco #Linux

Open source do GitHub: #15 a #11 | Vol. 12 #shorts Os projetos #15 a #11 do Top 15 open source do GitHub, Vol. 12, 10 segundos para cada um. #15 wanderer: open-wanderer/wanderer #14 mStream: IrosTheBeggar/mStream #13 Jumper: KingKongRobotics/jumper #12 withoutBG: withoutbg/withoutbg-python #11 Helical Insight: helicalinsight/helicalinsight Vídeo completo com os 15: youtu.be Oferecimento: Hallen Tech, hallentech.com #shorts #opensource #github

Mais 15 projetos open source do GitHub que valem o seu tempo | Vol. 12 Mais 15 projetos open source do GitHub que valem o seu tempo, em contagem regressiva e com 10 segundos para cada um: de um robô caranguejo a uma fábrica de software com agentes, passando pelo runtime da NVIDIA para isolar agentes de IA e um escritório em pixel art feito no Brasil. 📺 Décimo segundo volume da série Open source do GitHub. Escolhidos pela análise da lista inteira (estrelas, frequência de commits, último commit, README, documentação, capturas e site), conferida em 22/10/2026. Qual desses você vai testar primeiro? Conta nos comentários! ⏱️ CAPÍTULOS E LINKS 0:00 Abertura 0:10 #15 wanderer: catálogo self-hosted das suas trilhas de GPS, com mapas, planejamento de rotas e compartilhamento (4 mil estrelas) open-wanderer/wanderer 0:20 #14 mStream: servidor de música self-hosted com apps no celular, capas automáticas e acesso fora de casa (2,4 mil estrelas) IrosTheBeggar/mStream 0:30 #13 Jumper: robô caranguejo de 22 articulações e o kit de IA para criar aparência, movimentos e cenas (2,8 mil estrelas) KingKongRobotics/jumper 0:40 #12 withoutBG: remove o fundo de imagens em Python, localmente com pesos abertos ou pela API na nuvem (1,3 mil estrelas) withoutbg/withoutbg-python 0:50 #11 Helical Insight: plataforma de BI open source com dashboards, relatórios e chat analítico com IA (1,1 mil estrelas) helicalinsight/helicalinsight 1:00 #10 json-render: framework de UI generativa: a IA monta a interface só com os componentes do seu catálogo (18,8 mil estrelas) vercel-labs/json-render 1:10 #9 OpenShell: runtime da NVIDIA que isola agentes de IA e aplica regras do que cada um pode acessar (17,8 mil estrelas) NVIDIA/OpenShell 1:20 #8 inlark: cliente de e-mail desktop para Linux, Windows e Mac, sem assinatura e sem telemetria inlark/inlark 1:30 #7 morphicons: transforma qualquer ícone SVG em outro com animação de mola, sem dependências (2,9 mil estrelas) guillermolg00/morphicons 1:40 #6 OpenOats: transcreve reuniões localmente no Mac e sugere falas a partir das suas notas (2,6 mil estrelas) yazinsai/OpenOats 1:50 #5 impeccable: skill com 24 comandos e regras que tiram a cara de IA do design de frontend (85 mil estrelas) pbakaus/impeccable 2:00 #4 Habblaud: escritório virtual em pixel art para os agentes do Claude Code e do Codex (projeto brasileiro) marmottajr/habblaud 2:10 #3 explainroo: vídeos explicativos com narração feitos pelo seu agente de IA, renderizados no seu computador vincentsch/explainroo 2:20 #2 ClawRouter: roteador que escolhe o modelo de IA mais barato capaz de cada tarefa (6,6 mil estrelas) BlockRunAI/ClawRouter 2:30 Oferecimento Hallen Tech 2:44 #1 Fabro: fábrica de software open source: o processo vira um grafo, os agentes executam e você aprova onde importa (1,7 mil estrelas) fabro-sh/fabro 2:54 Encerramento 🧡 Oferecimento: Hallen Tech, liderança de tecnologia sob demanda, software, cloud e segurança para a sua empresa. Site: hallentech.com E-mail: contato@hallentech.com Instagram: instagram.com LinkedIn da Hallen Tech: linkedin.com LinkedIn do Eddwan: linkedin.com Se curtiu, deixa o like, segue o canal e ativa as notificações para a próxima lista. #opensource #github #programacao #ferramentas

Jumper Jumper is an open-source Python toolkit for a 22-DoF crab robot that lets AI coding agents design appearances and scenes, train motions with reinforcement learning, and package them for deployment to the real robot. KingKongRobotics/jumper

Jumper GitHub Overview: MuJoCo Simulation for a Quad-Legged Robot jumper by kingkongrobotics: kingkongrobotics/jumper Jumper is an open-source robotics framework centered on a neural-network-driven digital twin of a quad-legged, crab-like robot. This overview follows the relationship between the core simulation and reinforcement-learning engine and the companion jumper-design asset package. It covers the Python and Git foundation, isolated environments, Git LFS for dense 3D meshes, MuJoCo simulation, task discovery, a flat-terrain dry-run, PPO-style tripod training, checkpoint playback, and the path toward policy export or custom scenes. The rendered walkthrough depicts these concepts as an illustrated simulation workflow. It is an overview of the repository’s architecture and operating flow, not proof of a completed local installation, independent benchmark, or verified live-hardware deployment. TimeStamps: 0:00 Digital twin and the Jumper robotics framework 0:36 Python, Git, and architectural constraints 1:06 Core repository and isolated environment 1:58 jumper-design assets and Git LFS meshes 2:52 MuJoCo extras and simulation dependencies 3:22 Task discovery and the flat-terrain dry run 3:55 PPO training and checkpoint flow 4:30 Playback in the MuJoCo simulator 5:10 Policy export and custom obstacle-course direction 🤖 Quad-legged crab robot as a neural-network digital twin 🧩 Core engine, design assets, and Git LFS mesh handling 🧠 Reinforcement learning, tripod gait, and checkpoint playback 🛠️ MuJoCo simulation with possible export and scene workflows For Embodied AI researchers and robotics developers, Jumper offers a compact way to think about simulation, motion-policy training, visual assets, and deployment boundaries together. Treat the simulated gait and export path as the repository workflow shown here, then consult the project documentation before attempting any physical integration. #JumperRobot #MuJoCo #ReinforcementLearning [de-DE] Jumper GitHub-Überblick: MuJoCo-Simulation für einen vierbeinigen Roboter Jumper von Kingkongrobotics: kingkongrobotics/jumper Jumper ist ein Open-Source-Robotik-Framework, das auf einem neuronalen Netzwerk basierenden digitalen Zwilling eines vierbeinigen, krabbenartigen Roboters basiert. Diese Übersicht beschreibt die Beziehung zwischen der Kernsimulations- und Reinforcement-Learning-Engine und dem zugehörigen Jumper-Design-Asset-Paket. Sie behandelt die Python- und Git-Grundlagen, isolierte Umgebungen, Git LFS für dichte 3D-Netze, MuJoCo-Simulation, Aufgabenerkennung, einen Testlauf auf flachem Gelände, PPO-ähnliches Tripod-Training, Checkpoint-Wiedergabe und den Weg zum Richtlinienexport oder zur Erstellung benutzerdefinierter Szenen. Die gerenderte Demonstration veranschaulicht diese Konzepte als illustrierten Simulations-Workflow. Sie bietet einen Überblick über die Architektur und den Betriebsablauf des Repositorys und stellt keinen Nachweis für eine abgeschlossene lokale Installation, einen unabhängigen Benchmark oder eine verifizierte Live-Hardware-Bereitstellung dar. Zeitstempel: 0:00 Digitaler Zwilling und das Jumper-Robotik-Framework 0:36 Python, Git und architektonische Einschränkungen 1:06 Kern-Repository und isolierte Umgebung 1:58 Jumper-Design-Assets und Git-LFS-Meshes 2:52 MuJoCo-Extras und Simulationsabhängigkeiten 3:22 Aufgabenerkennung und Testlauf auf ebener Fläche 3:55 PPO-Training und Checkpoint-Ablauf 4:30 Wiedergabe im MuJoCo-Simulator 5:10 Richtlinienexport und benutzerdefinierte Hindernisparcours-Richtung 🤖 Vierbeiniger Krabbenroboter als neuronaler Netzwerk-Digitalzwilling 🧩 Kern-Engine, Design-Assets und Git-LFS-Mesh-Handling 🧠 Reinforcement Learning, Tripod-Gang und Checkpoint-Wiedergabe 🛠️ MuJoCo-Simulation mit möglichen Export- und Szenen-Workflows Für Forscher im Bereich Embodied AI und Robotik-Entwickler bietet Jumper eine kompakte Möglichkeit, Simulation, Bewegungsrichtlinien-Training, visuelle Assets und Einsatzgrenzen gemeinsam zu betrachten. Behandeln Sie den simulierten Gang und den Exportpfad gemäß dem hier gezeigten Repository-Workflow und konsultieren Sie anschließend die Projektdokumentation, bevor Sie eine physische Integration vornehmen. #JumperRobot #MuJoCo #ReinforcementLearning [es-US] Visión general de Jumper en GitHub: Simulación en MuJoCo para un robot de cuatro patas Jumper de kingkongrobotics: kingkongrobotics/jumper Jumper es un marco de robótica de código abierto centrado en un gemelo digital de un robot cuadrúpedo con forma de cangrejo, impulsado por redes neuronales. Esta descripción general analiza la relación entre el motor principal de simulación y aprendizaje por refuerzo y el paquete de recursos complementarios para el diseño de Jumper. Cubre los fundamentos de Python y Git, entornos aislados, Git LFS para mallas 3D densas, simulación con MuJoCo, descubrimiento de tareas, una prueba en terreno plano, entrenamiento con trípode estilo PPO, reproducción de puntos de control y el camino hacia la exportación de políticas o escenas personalizadas. El recorrido virtual ilustra estos conceptos como un flujo de trabajo de simulación. Se trata de una descripción general de la arquitectura y el flujo operativo del repositorio, no una prueba de una instalación local completa, una evaluación comparativa independiente ni una implementación verificada en hardware real. Marcas de tiempo: 0:00 Gemelo digital y el marco de robótica Jumper 0:36 Python, Git y restricciones arquitectónicas 1:06 Repositorio principal y entorno aislado 1:58 Recursos de diseño de Jumper y mallas Git LFS 2:52 Complementos de MuJoCo y dependencias de simulación 3:22 Descubrimiento de tareas y prueba en terreno plano 3:55 Entrenamiento de PPO y flujo de puntos de control 4:30 Reproducción en el simulador MuJoCo 5:10 Exportación de políticas y dirección personalizada de la pista de obstáculos 🤖 Robot cangrejo cuadrúpedo como gemelo digital de red neuronal 🧩 Motor principal, recursos de diseño y manejo de mallas Git LFS 🧠 Aprendizaje por refuerzo, marcha trípode y reproducción de puntos de control 🛠️ Simulación MuJoCo con posibles flujos de trabajo de exportación y escena Para investigadores de IA integrada y desarrolladores de robótica, Jumper ofrece una forma compacta de integrar la simulación, el entrenamiento de políticas de movimiento, los recursos visuales y los límites de implementación. Trata la marcha simulada y la ruta de exportación como el flujo de trabajo del repositorio que se muestra aquí, y luego consulta la documentación del proyecto antes de intentar cualquier integración física. #RobotSaltador #MuJoCo #AprendizajePorRefuerzo [fr-FR] Présentation de Jumper sur GitHub : Simulation MuJoCo pour robot quadrupède Jumper par Kingkongrobotics : kingkongrobotics/jumper Jumper est un framework robotique open source basé sur un jumeau numérique, piloté par un réseau neuronal, d'un robot quadrupède ressemblant à un crabe. Cette présentation décrit la relation entre le moteur de simulation et d'apprentissage par renforcement et le package d'assets jumper-design. Elle aborde les bases Python et Git, les environnements isolés, Git LFS pour les maillages 3D denses, la simulation MuJoCo, la découverte de tâches, une simulation à blanc sur terrain plat, l'entraînement sur trépied de type PPO, la lecture des points de contrôle et la possibilité d'exporter des politiques ou de créer des scènes personnalisées. La visite guidée illustrée présente ces concepts sous forme de flux de travail de simulation. Il s'agit d'une présentation de l'architecture et du fonctionnement du dépôt, et non d'une preuve d'installation locale complète, d'un benchmark indépendant ou d'un déploiement matériel réel vérifié. Horodatage : 0:00 Jumeau numérique et framework robotique Jumper 0:36 Python, Git et contraintes architecturales 1:06 Dépôt principal et environnement isolé 1:58 Ressources de conception Jumper et maillages Git LFS 2:52 Fonctionnalités MuJoCo et dépendances de simulation 3:22 Découverte des tâches et simulation sur terrain plat 3:55 Entraînement PPO et flux de points de contrôle 4:30 Lecture dans le simulateur MuJoCo 5:10 Exportation de politiques et direction personnalisée du parcours d'obstacles 🤖 Robot crabe quadrupède : un jumeau numérique à réseau neuronal 🧩 Moteur principal, ressources de conception et gestion des maillages Git LFS 🧠 Apprentissage par renforcement, démarche tripode et lecture des points de contrôle 🛠️ Simulation MuJoCo avec possibilités d'exportation et de flux de travail de scènes Pour les chercheurs en IA incarnée et les développeurs en robotique, Jumper offre une solution compacte pour appréhender simultanément la simulation, l'entraînement des politiques de mouvement, les ressources visuelles et les limites de déploiement. Considérez la démarche simulée et la trajectoire d'exportation comme le flux de travail du référentiel présenté ici, puis consultez la documentation du projet avant toute intégration physique. #JumperRobot #MuJoCo #ApprentissageParRenforcement [hi] Jumper GitHub का अवलोकन: चार पैरों वाले रोबोट के लिए MuJoCo सिमुलेशन जम्पर (Jumper) किंगकोंग्रोबोटिक्स द्वारा: kingkongrobotics/jumper जम्पर एक ओपन-सोर्स रोबोटिक्स फ्रेमवर्क है जो चार पैरों वाले, केकड़े जैसे रोबोट के न्यूरल नेटवर्क-आधारित डिजिटल ट्विन पर केंद्रित है। यह अवलोकन कोर सिमुलेशन और रीइन्फोर्समेंट-लर्निंग इंजन तथा इसके साथ आने वाले जम्पर-डिज़ाइन एसेट पैकेज के बीच संबंधों को दर्शाता है। इसमें पायथन और गिट की बुनियादी बातें, आइसोलेटेड एनवायरनमेंट, सघन 3डी मेश के लिए गिट एलएफएस, मुजोको सिमुलेशन, टास्क डिस्कवरी, समतल भूभाग पर ड्राई-रन, पीपीओ-शैली ट्राइपॉड ट्रेनिंग, चेकपॉइंट प्लेबैक और पॉलिसी एक्सपोर्ट या कस्टम सीन बनाने की प्रक्रिया शामिल है। चित्रित वॉकथ्रू इन अवधारणाओं को एक सचित्र सिमुलेशन वर्कफ़्लो के रूप में दर्शाता है। यह रिपॉजिटरी की संरचना और संचालन प्रवाह का एक अवलोकन है, न कि पूर्ण स्थानीय इंस्टॉलेशन, स्वतंत्र बेंचमार्क या सत्यापित लाइव हार्डवेयर परिनियोजन का प्रमाण। टाइमस्टैम्प: 0:00 डिजिटल ट्विन और जम्पर रोबोटिक्स फ्रेमवर्क 0:36 पायथन, गिट और आर्किटेक्चरल बाधाएँ 1:06 कोर रिपॉजिटरी और पृथक वातावरण 1:58 जम्पर-डिज़ाइन एसेट्स और गिट एलएफएस मेश 2:52 मुजोको एक्स्ट्रा और सिमुलेशन निर्भरताएँ 3:22 टास्क डिस्कवरी और समतल भूभाग पर ड्राई रन 3:55 पीपीओ प्रशिक्षण और चेकपॉइंट प्रवाह 4:30 मुजोको सिम्युलेटर में प्लेबैक 5:10 पॉलिसी निर्यात और कस्टम बाधा-मार्ग दिशा 🤖 चार पैरों वाला केकड़ा रोबोट एक न्यूरल-नेटवर्क डिजिटल ट्विन के रूप में 🧩 कोर इंजन, डिज़ाइन एसेट्स और गिट एलएफएस मेश हैंडलिंग 🧠 रीइन्फोर्समेंट लर्निंग, ट्राइपॉड चाल और चेकपॉइंट प्लेबैक 🛠️ संभावित निर्यात और दृश्य वर्कफ़्लो के साथ मुजोको सिमुलेशन एम्बोडेड एआई शोधकर्ताओं और रोबोटिक्स डेवलपर्स के लिए, जम्पर सिमुलेशन, मोशन-पॉलिसी प्रशिक्षण, विज़ुअल एसेट्स और परिनियोजन सीमाओं के बारे में एक साथ सोचने का एक संक्षिप्त तरीका प्रदान करता है। सिम्युलेटेड चाल और निर्यात पथ को यहाँ दिखाए गए रिपॉजिटरी वर्कफ़्लो के रूप में मानें, और किसी भी भौतिक एकीकरण का प्रयास करने से पहले प्रोजेक्ट दस्तावेज़ देखें। #JumperRobot #MuJoCo #ReinforcementLearning [id] Ikhtisar GitHub Jumper: Simulasi MuJoCo untuk Robot Berkaki Empat Jumper oleh KingkongRobotics: kingkongRobotics/jumper Jumper adalah kerangka kerja robotika sumber terbuka yang berpusat pada kembaran digital berbasis jaringan saraf dari robot berkaki empat mirip kepiting. Gambaran umum ini mengikuti hubungan antara mesin simulasi inti dan pembelajaran penguatan serta paket aset desain Jumper yang menyertainya. Ini mencakup fondasi Python dan Git, lingkungan terisolasi, Git LFS untuk mesh 3D padat, simulasi MuJoCo, penemuan tugas, uji coba medan datar, pelatihan tripod gaya PPO, pemutaran titik pemeriksaan, dan jalur menuju ekspor kebijakan atau adegan khusus. Penjelasan yang dirender menggambarkan konsep-konsep ini sebagai alur kerja simulasi yang diilustrasikan. Ini adalah gambaran umum arsitektur dan alur operasi repositori, bukan bukti instalasi lokal yang lengkap, tolok ukur independen, atau penerapan perangkat keras langsung yang terverifikasi. Cap Waktu: 0:00 Kembaran digital dan kerangka kerja robotika Jumper 0:36 Python, Git, dan batasan arsitektur 1:06 Repositori inti dan lingkungan terisolasi 1:58 Aset desain jumper dan mesh Git LFS 2:52 Ekstra MuJoCo dan dependensi simulasi 3:22 Penemuan tugas dan uji coba medan datar 3:55 Pelatihan PPO dan alur titik pemeriksaan 4:30 Pemutaran di simulator MuJoCo 5:10 Ekspor kebijakan dan arah rintangan khusus 🤖 Robot kepiting berkaki empat sebagai kembaran digital jaringan saraf 🧩 Mesin inti, aset desain, dan penanganan mesh Git LFS 🧠 Pembelajaran penguatan, langkah tripod, dan pemutaran titik pemeriksaan 🛠️ Simulasi MuJoCo dengan kemungkinan ekspor dan alur kerja adegan Bagi peneliti AI yang diwujudkan dan pengembang robotika, Jumper menawarkan cara ringkas untuk memikirkan simulasi, pelatihan kebijakan gerak, aset visual, dan batasan penyebaran secara bersamaan. Anggap simulasi gerakan dan jalur ekspor sebagai alur kerja repositori yang ditunjukkan di sini, lalu konsultasikan dokumentasi proyek sebelum mencoba integrasi fisik apa pun. #JumperRobot #MuJoCo #ReinforcementLearning [it] Panoramica di Jumper su GitHub: Simulazione MuJoCo per un robot a quattro zampe Jumper di kingkongrobotics: kingkongrobotics/jumper Jumper è un framework robotico open-source incentrato su un gemello digitale, basato su reti neurali, di un robot quadrupede simile a un granchio. Questa panoramica illustra la relazione tra il motore di simulazione e apprendimento per rinforzo e il pacchetto di risorse per la progettazione di Jumper. Vengono trattati i fondamenti di Python e Git, gli ambienti isolati, Git LFS per mesh 3D dense, la simulazione con MuJoCo, l'individuazione dei compiti, una simulazione su terreno pianeggiante, l'addestramento con treppiede in stile PPO, la riproduzione dei checkpoint e il percorso verso l'esportazione delle policy o la creazione di scene personalizzate. La procedura guidata renderizzata illustra questi concetti attraverso un flusso di lavoro di simulazione. Si tratta di una panoramica dell'architettura e del flusso operativo del repository, non di una dimostrazione di un'installazione locale completa, di un benchmark indipendente o di un'implementazione verificata su hardware reale. Indice dei contenuti: 0:00 Gemello digitale e framework robotico Jumper 0:36 Python, Git e vincoli architetturali 1:06 Repository principale e ambiente isolato 1:58 Risorse di progettazione di Jumper e mesh Git LFS 2:52 Funzionalità aggiuntive di MuJoCo e dipendenze di simulazione 3:22 Individuazione dei task e simulazione su terreno pianeggiante 3:55 Addestramento PPO e flusso di checkpoint 4:30 Riproduzione nel simulatore MuJoCo 5:10 Esportazione delle policy e direzione personalizzata del percorso a ostacoli 🤖 Robot granchio a quattro zampe come gemello digitale di una rete neurale 🧩 Motore principale, risorse di progettazione e gestione delle mesh Git LFS 🧠 Apprendimento per rinforzo, andatura a treppiede e riproduzione dei checkpoint 🛠️ Simulazione MuJoCo con possibili flussi di lavoro di esportazione e scene Per i ricercatori di IA incarnata e gli sviluppatori di robotica, Jumper offre un modo compatto per pensare insieme a simulazione, addestramento delle policy di movimento, risorse visive e limiti di implementazione. Considerate la simulazione dell'andatura e il percorso di esportazione come il flusso di lavoro del repository mostrato qui, quindi consultate la documentazione del progetto prima di tentare qualsiasi integrazione fisica. #JumperRobot #MuJoCo #ReinforcementLearning [pt-BR] Visão Geral do GitHub do Jumper: Simulação MuJoCo para um Robô Quadrúpede Jumper, da Kingkongrobotics: kingkongrobotics/jumper Jumper é um framework de robótica de código aberto centrado em um gêmeo digital, baseado em redes neurais, de um robô quadrúpede semelhante a um caranguejo. Esta visão geral aborda a relação entre o núcleo de simulação e o mecanismo de aprendizado por reforço, bem como o pacote de recursos de design do Jumper. Abrange os fundamentos em Python e Git, ambientes isolados, Git LFS para malhas 3D densas, simulação MuJoCo, descoberta de tarefas, um teste em terreno plano, treinamento com tripé no estilo PPO, reprodução de checkpoints e o caminho para exportação de políticas ou cenas personalizadas. O passo a passo renderizado ilustra esses conceitos como um fluxo de trabalho de simulação. Trata-se de uma visão geral da arquitetura e do fluxo operacional do repositório, e não de uma comprovação de instalação local completa, benchmark independente ou implantação verificada em hardware real. Marcadores de tempo: 0:00 Gêmeo digital e a estrutura robótica Jumper 0:36 Python, Git e restrições arquitetônicas 1:06 Repositório principal e ambiente isolado 1:58 Recursos de design do Jumper e malhas Git LFS 2:52 Extras do MuJoCo e dependências de simulação 3:22 Descoberta de tarefas e o teste em terreno plano 3:55 Treinamento de PPO e fluxo de checkpoints 4:30 Reprodução no simulador MuJoCo 5:10 Exportação de políticas e direção personalizada de percurso com obstáculos 🤖 Robô caranguejo quadrúpede como um gêmeo digital de rede neural 🧩 Motor principal, recursos de design e manipulação de malhas Git LFS 🧠 Aprendizado por reforço, marcha em tripé e reprodução de checkpoints 🛠️ Simulação MuJoCo com possíveis fluxos de trabalho de exportação e cena Para pesquisadores de IA incorporada e desenvolvedores de robótica, o Jumper oferece uma maneira compacta de pensar sobre simulação, treinamento de políticas de movimento, recursos visuais e limites de implantação em conjunto. Considere a marcha simulada e o caminho de exportação como o fluxo de trabalho do repositório mostrado aqui e, em seguida, consulte a documentação do projeto antes de tentar qualquer integração física. #JumperRobot #MuJoCo #ReinforcementLearning

@AlertaNews24 Se llama Jumper, aquí el repositorio: KingKongRobotics/jumper

🦀Jumper is open source on GitHub: KingKongRobotics/jumper

兄弟们，开源的小螃蟹机器人 Jumper真的太酷啦！它拥有22个高自由度动作。不但长得萌，动作还很灵活！能搬东西，甚至能跳舞。 现在作者已经把整个项目开源了，想要的兄弟自己就能照着搭一只！ 传送门：KingKongRobotics/jumper t.co

結構デカイのでちくわがびびりそう KingKongRobotics/jumper