{"id":"92d6e8dd7a0c6465d0a96732aaa0d186db724fc30fd7cf9de2ed2459af062fbd","title":"Authors present Crazyflow, a JAX drone simulator for fast, differentiable training","argument":"The authors of Crazyflow present a drone simulator built with JAX, a software framework, that combines high-fidelity dynamics with GPU acceleration and differentiability. They say it supports model-based and data-driven methods, gradient-based and sampling-based approaches, and systems ranging from single drones to swarms.\n\nThe authors report scaling to millions of environments or thousands of swarms with thousands of drones on single CPUs and GPUs, and beyond on clusters. In real-world experiments, they report analytical-gradient-based policy learning for high-accuracy tracking without domain randomization, and sampling-based obstacle avoidance at speeds exceeding half a billion steps per second.\n\nThey also describe an in-flight reinforcement learning demonstration: a physical drone was thrown into the air, trained to recover from scratch in 0.37 seconds, and caught before hitting the ground. This suggests a possible way to train or optimize a recovery policy during flight, rather than only training before deployment; the abstract does not establish how broadly this approach works or how it compares with other methods.\n\nHow does the recovery demonstration perform across repeated trials and different flight conditions?","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The supplied abstract does not describe the methods, trial counts, flight conditions, or comparison baselines for the reported real-world experiments. The abstract does not establish how broadly in-flight reinforcement learning works beyond the described recovery demonstration.","speaker":{"name":"Schuck, Martin, Rath, Marcel P., Hua, Yufei, Goudar, Abhishek, Zhou, SiQi, Schoellig, Angela P.","roleAtSource":"Article author"},"source":{"title":"Crazyflow: An Accurate, GPU-Accelerated, Differentiable Drone Simulator in JAX","publisher":"arXiv eess.SY","format":"Article","url":"https://arxiv.org/abs/2606.01478","publishedAt":"2026-05-31","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors present Crazyflow as a drone simulator combining high-fidelity dynamics with JAX for speed and differentiability.","excerpt":"Crazyflow combines high-fidelity dynamics with JAX for speed and differentiability","url":"https://arxiv.org/abs/2606.01478","relationship":"Source-stated"},{"text":"The authors report that Crazyflow scales to millions of environments or thousands of swarms with thousands of drones on single CPUs and GPUs, and beyond on clusters.","excerpt":"scaling to millions of environments or thousands of swarms with thousands of drones on single CPUs and GPUs, and beyond on clusters","url":"https://arxiv.org/abs/2606.01478","relationship":"Source-stated"},{"text":"The authors report real-world experiments involving tracking without domain randomization and sampling-based obstacle avoidance at speeds exceeding half a billion steps per second.","excerpt":"Real-world experiments show Crazyflow supports both analytical-gradient-based policy learning, achieving high-accuracy tracking without domain randomization, and sampling-based obstacle avoidance at speeds exceeding half a billion steps per second.","url":"https://arxiv.org/abs/2606.01478","relationship":"Source-stated"},{"text":"The authors say they trained a recovery policy from scratch in 0.37 seconds after throwing a physical drone into the air, and caught it before it hit the ground.","excerpt":"throwing a physical drone into the air and training a recovery policy from scratch in 0.37 seconds, catching it before it hits the ground","url":"https://arxiv.org/abs/2606.01478","relationship":"Source-stated"},{"text":"The demonstration suggests a possible route to training a recovery policy during flight, but the abstract does not establish how broadly it works or how it compares with other methods.","excerpt":"Crazyflow's performance enables in-flight reinforcement learning; we demonstrate this by throwing a physical drone into the air and training a recovery policy from scratch in 0.37 seconds","url":"https://arxiv.org/abs/2606.01478","relationship":"Escape Meridian interpretation"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2606.01478","checkedAt":"2026-11-06T01:20:06.804Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How does the recovery demonstration perform across repeated trials and different flight conditions?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Authors present Crazyflow, a JAX drone simulator for fast, differentiable training","summary":"The authors of Crazyflow present a drone simulator built with JAX, a software framework, that combines high-fidelity dynamics with GPU acceleration and differentiability. They say it supports model-based and data-driven methods, gradient-based and sampling-based approaches, and systems ranging from single drones to swarms.\n\nThe authors report scaling to millions of environments or thousands of swarms with thousands of drones on single CPUs and GPUs, and beyond on clusters. In real-world experiments, they report analytical-gradient-based policy learning for high-accuracy tracking without domain randomization, and sampling-based obstacle avoidance at speeds exceeding half a billion steps per second.\n\nThey also describe an in-flight reinforcement learning demonstration: a physical drone was thrown into the air, trained to recover from scratch in 0.37 seconds, and caught before hitting the ground. This suggests a possible way to train or optimize a recovery policy during flight, rather than only training before deployment; the abstract does not establish how broadly this approach works or how it compares with other methods.\n\nHow does the recovery demonstration perform across repeated trials and different flight conditions?","question":"How does the recovery demonstration perform across repeated trials and different flight conditions?","claims":[{"text":"The authors present Crazyflow as a drone simulator combining high-fidelity dynamics with JAX for speed and differentiability.","excerpt":"Crazyflow combines high-fidelity dynamics with JAX for speed and differentiability","support":"source-stated"},{"text":"The authors report that Crazyflow scales to millions of environments or thousands of swarms with thousands of drones on single CPUs and GPUs, and beyond on clusters.","excerpt":"scaling to millions of environments or thousands of swarms with thousands of drones on single CPUs and GPUs, and beyond on clusters","support":"source-stated"},{"text":"The authors report real-world experiments involving tracking without domain randomization and sampling-based obstacle avoidance at speeds exceeding half a billion steps per second.","excerpt":"Real-world experiments show Crazyflow supports both analytical-gradient-based policy learning, achieving high-accuracy tracking without domain randomization, and sampling-based obstacle avoidance at speeds exceeding half a billion steps per second.","support":"source-stated"},{"text":"The authors say they trained a recovery policy from scratch in 0.37 seconds after throwing a physical drone into the air, and caught it before it hit the ground.","excerpt":"throwing a physical drone into the air and training a recovery policy from scratch in 0.37 seconds, catching it before it hits the ground","support":"source-stated"},{"text":"The demonstration suggests a possible route to training a recovery policy during flight, but the abstract does not establish how broadly it works or how it compares with other methods.","excerpt":"Crazyflow's performance enables in-flight reinforcement learning; we demonstrate this by throwing a physical drone into the air and training a recovery policy from scratch in 0.37 seconds","support":"interpretation"}],"limitations":["The supplied abstract does not describe the methods, trial counts, flight conditions, or comparison baselines for the reported real-world experiments.","The abstract does not establish how broadly in-flight reinforcement learning works beyond the described recovery demonstration."],"mediaId":"9d5272325537e4d141361915b16baffcdc73d17d6f976a173a8dc8c495228133","correctionReason":null,"id":"92d6e8dd7a0c6465d0a96732aaa0d186db724fc30fd7cf9de2ed2459af062fbd","publicationId":"bc5a282d-44c6-4a13-89aa-57b5e86053e9","version":1,"sourceUrl":"https://arxiv.org/abs/2606.01478","sourceTitle":"Crazyflow: An Accurate, GPU-Accelerated, Differentiable Drone Simulator in JAX","publisher":"arXiv eess.SY","publishedAt":null,"originalPublished":{"value":"2026-05-31","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-10-27T04:31:01.286Z","createdAt":"2026-11-06T00:17:21.759Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"9d5272325537e4d141361915b16baffcdc73d17d6f976a173a8dc8c495228133","kind":"article","url":"https://arxiv.org/abs/2606.01478","sourceUrl":"https://arxiv.org/abs/2606.01478","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-10-30T00:31:31.033Z","method":"acquired_source","contentType":"text/html"},"recommendationReason":"Recommended by Jev for understanding this source; checked by Luna."}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2606.01478","checkedAt":"2026-11-06T01:20:06.804Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"9d5272325537e4d141361915b16baffcdc73d17d6f976a173a8dc8c495228133","confidence":0.85,"provider":"typesafe","model":"jev-1.13.0","inputHash":"1261460e137e42ddb073b3a028c0ade98d5504107306325834e03574e44e9bb9","decidedAt":"2026-10-30T00:31:34.440Z"},"provenance":{"textHash":"aefb0240e1d6441deb627fa7b341b23aded1660b8dced653729111b29314401e","observationId":"535f178817c61c42ef7d288a62ab96a437f6d56c070d41276e67437fb76e0145","retrievedAt":"2026-10-30T00:31:30.443Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"6f7ce201-3ef6-480d-807c-4ade8da5ab5a"},"reviewer":{"model":"gpt-6-luna","runId":"ebaea321-bad0-4bd8-8d88-75840047ee9b"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-06T00:17:21.759Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Schuck, Martin","url":null,"kind":"unknown"},{"name":"Rath, Marcel P.","url":null,"kind":"unknown"},{"name":"Hua, Yufei","url":null,"kind":"unknown"},{"name":"Goudar, Abhishek","url":null,"kind":"unknown"},{"name":"Zhou, SiQi","url":null,"kind":"unknown"},{"name":"Schoellig, Angela P.","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Schuck, Martin","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Schuck, Martin","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2606.01478","locator":"head/meta[name=author|citation_author]"},"assertionId":"114659ef6fa25fa15bc8b5c82e2da38f039a60f5425ceb8a625ab41b555de4fd:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-06T01:20:06.804Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"114659ef6fa25fa15bc8b5c82e2da38f039a60f5425ceb8a625ab41b555de4fd","sourceUrl":"https://arxiv.org/abs/2606.01478","contentSha256":"8e944ad844c11a22eac35bc943b7580d5c424ffa3cf9dd5bed746b2c9dba7325","textHash":"aefb0240e1d6441deb627fa7b341b23aded1660b8dced653729111b29314401e","observedAt":"2026-11-06T01:20:06.804Z"}},{"actor":{"name":"Rath, Marcel P.","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Rath, Marcel P.","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2606.01478","locator":"head/meta[name=author|citation_author]"},"assertionId":"114659ef6fa25fa15bc8b5c82e2da38f039a60f5425ceb8a625ab41b555de4fd:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-06T01:20:06.804Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"114659ef6fa25fa15bc8b5c82e2da38f039a60f5425ceb8a625ab41b555de4fd","sourceUrl":"https://arxiv.org/abs/2606.01478","contentSha256":"8e944ad844c11a22eac35bc943b7580d5c424ffa3cf9dd5bed746b2c9dba7325","textHash":"aefb0240e1d6441deb627fa7b341b23aded1660b8dced653729111b29314401e","observedAt":"2026-11-06T01:20:06.804Z"}},{"actor":{"name":"Hua, Yufei","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Hua, Yufei","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2606.01478","locator":"head/meta[name=author|citation_author]"},"assertionId":"114659ef6fa25fa15bc8b5c82e2da38f039a60f5425ceb8a625ab41b555de4fd:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-06T01:20:06.804Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"114659ef6fa25fa15bc8b5c82e2da38f039a60f5425ceb8a625ab41b555de4fd","sourceUrl":"https://arxiv.org/abs/2606.01478","contentSha256":"8e944ad844c11a22eac35bc943b7580d5c424ffa3cf9dd5bed746b2c9dba7325","textHash":"aefb0240e1d6441deb627fa7b341b23aded1660b8dced653729111b29314401e","observedAt":"2026-11-06T01:20:06.804Z"}},{"actor":{"name":"Goudar, Abhishek","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Goudar, Abhishek","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2606.01478","locator":"head/meta[name=author|citation_author]"},"assertionId":"114659ef6fa25fa15bc8b5c82e2da38f039a60f5425ceb8a625ab41b555de4fd:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-06T01:20:06.804Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"114659ef6fa25fa15bc8b5c82e2da38f039a60f5425ceb8a625ab41b555de4fd","sourceUrl":"https://arxiv.org/abs/2606.01478","contentSha256":"8e944ad844c11a22eac35bc943b7580d5c424ffa3cf9dd5bed746b2c9dba7325","textHash":"aefb0240e1d6441deb627fa7b341b23aded1660b8dced653729111b29314401e","observedAt":"2026-11-06T01:20:06.804Z"}},{"actor":{"name":"Zhou, SiQi","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Zhou, SiQi","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2606.01478","locator":"head/meta[name=author|citation_author]"},"assertionId":"114659ef6fa25fa15bc8b5c82e2da38f039a60f5425ceb8a625ab41b555de4fd:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-06T01:20:06.804Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"114659ef6fa25fa15bc8b5c82e2da38f039a60f5425ceb8a625ab41b555de4fd","sourceUrl":"https://arxiv.org/abs/2606.01478","contentSha256":"8e944ad844c11a22eac35bc943b7580d5c424ffa3cf9dd5bed746b2c9dba7325","textHash":"aefb0240e1d6441deb627fa7b341b23aded1660b8dced653729111b29314401e","observedAt":"2026-11-06T01:20:06.804Z"}},{"actor":{"name":"Schoellig, Angela P.","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Schoellig, Angela P.","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2606.01478","locator":"head/meta[name=author|citation_author]"},"assertionId":"114659ef6fa25fa15bc8b5c82e2da38f039a60f5425ceb8a625ab41b555de4fd:participation:5","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-06T01:20:06.804Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"114659ef6fa25fa15bc8b5c82e2da38f039a60f5425ceb8a625ab41b555de4fd","sourceUrl":"https://arxiv.org/abs/2606.01478","contentSha256":"8e944ad844c11a22eac35bc943b7580d5c424ffa3cf9dd5bed746b2c9dba7325","textHash":"aefb0240e1d6441deb627fa7b341b23aded1660b8dced653729111b29314401e","observedAt":"2026-11-06T01:20:06.804Z"}}],"participationCoverage":{"total":6,"returned":6,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2606.01478","snapshotId":"114659ef6fa25fa15bc8b5c82e2da38f039a60f5425ceb8a625ab41b555de4fd","contentSha256":"8e944ad844c11a22eac35bc943b7580d5c424ffa3cf9dd5bed746b2c9dba7325","textHash":"aefb0240e1d6441deb627fa7b341b23aded1660b8dced653729111b29314401e","observedAt":"2026-11-06T01:20:06.804Z","basis":"publisher-page-metadata","sourceLinks":52,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/92d6e8dd7a0c6465d0a96732aaa0d186db724fc30fd7cf9de2ed2459af062fbd/source"}},"canonicalUrl":"https://escapemeridian.com/readings/authors-present-crazyflow-a-jax-drone-simulator-for-fast-92d6e8dd","formats":{"html":"/readings/authors-present-crazyflow-a-jax-drone-simulator-for-fast-92d6e8dd","markdown":"/readings/authors-present-crazyflow-a-jax-drone-simulator-for-fast-92d6e8dd.md","json":"/readings/authors-present-crazyflow-a-jax-drone-simulator-for-fast-92d6e8dd.json"},"attribution":"Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.","maintainedGuideLinks":[],"readerContext":{"version":"reader-context-v1","attribution":"Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.","terms":[{"id":"term-gpu","name":"Graphics processing unit","explanation":"A chip that performs many calculations at once; the main hardware used to train and run AI models.","referenceUrl":"https://en.wikipedia.org/wiki/Graphics_processing_unit"},{"id":"term-cpu","name":"CPU","explanation":"The central processing unit: a general-purpose processor that carries out a computer's instructions.","referenceUrl":"https://aws.amazon.com/what-is/cpu/"},{"id":"term-reinforcement-learning","name":"Reinforcement learning","explanation":"Training by trial and error, in which a system is rewarded for the outcomes its designers want.","referenceUrl":"https://developers.google.com/machine-learning/glossary#reinforcement-learning-rl"}],"entities":[{"id":"arxiv","name":"arXiv","kind":"organization","context":"arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.","referenceUrls":["https://arxiv.org/","https://info.arxiv.org/help/api/index.html"],"path":"/entities/arxiv","json":"/entities/arxiv.json","markdown":"/entities/arxiv.md"}]},"researchQuestions":[],"withdrawn":false,"references":{"schemaVersion":1,"target":{"type":"reading","id":"92d6e8dd7a0c6465d0a96732aaa0d186db724fc30fd7cf9de2ed2459af062fbd","revision":"bc5a282d-44c6-4a13-89aa-57b5e86053e9","revisionKind":"publication-id","title":"Authors present Crazyflow, a JAX drone simulator for fast, differentiable training","url":"https://escapemeridian.com/readings/authors-present-crazyflow-a-jax-drone-simulator-for-fast-92d6e8dd"},"source":{"type":"source","id":"https://arxiv.org/abs/2606.01478","revision":"aefb0240e1d6441deb627fa7b341b23aded1660b8dced653729111b29314401e","revisionKind":"sha256-source-text","title":"Crazyflow: An Accurate, GPU-Accelerated, Differentiable Drone Simulator in JAX","url":"https://arxiv.org/abs/2606.01478"},"status":"available","items":[],"next":null,"scope":"Explicit citations in public readings and public file attachments to this reading. Shared topics are not citations. Other artifact types need an explicit reference adapter."},"revisions":[{"id":"92d6e8dd7a0c6465d0a96732aaa0d186db724fc30fd7cf9de2ed2459af062fbd","title":"Authors present Crazyflow, a JAX drone simulator for fast, differentiable training","argument":"The authors of Crazyflow present a drone simulator built with JAX, a software framework, that combines high-fidelity dynamics with GPU acceleration and differentiability. They say it supports model-based and data-driven methods, gradient-based and sampling-based approaches, and systems ranging from single drones to swarms.\n\nThe authors report scaling to millions of environments or thousands of swarms with thousands of drones on single CPUs and GPUs, and beyond on clusters. In real-world experiments, they report analytical-gradient-based policy learning for high-accuracy tracking without domain randomization, and sampling-based obstacle avoidance at speeds exceeding half a billion steps per second.\n\nThey also describe an in-flight reinforcement learning demonstration: a physical drone was thrown into the air, trained to recover from scratch in 0.37 seconds, and caught before hitting the ground. This suggests a possible way to train or optimize a recovery policy during flight, rather than only training before deployment; the abstract does not establish how broadly this approach works or how it compares with other methods.\n\nHow does the recovery demonstration perform across repeated trials and different flight conditions?","topicIds":["intelligence-and-discovery"],"domain":"science","claimType":"attributed reading","limits":"The supplied abstract does not describe the methods, trial counts, flight conditions, or comparison baselines for the reported real-world experiments. The abstract does not establish how broadly in-flight reinforcement learning works beyond the described recovery demonstration.","speaker":{"name":"Schuck, Martin, Rath, Marcel P., Hua, Yufei, Goudar, Abhishek, Zhou, SiQi, Schoellig, Angela P.","roleAtSource":"Article author"},"source":{"title":"Crazyflow: An Accurate, GPU-Accelerated, Differentiable Drone Simulator in JAX","publisher":"arXiv eess.SY","format":"Article","url":"https://arxiv.org/abs/2606.01478","publishedAt":"2026-05-31","dateBasis":"Date basis not provided."},"evidence":[{"text":"The authors present Crazyflow as a drone simulator combining high-fidelity dynamics with JAX for speed and differentiability.","excerpt":"Crazyflow combines high-fidelity dynamics with JAX for speed and differentiability","url":"https://arxiv.org/abs/2606.01478","relationship":"Source-stated"},{"text":"The authors report that Crazyflow scales to millions of environments or thousands of swarms with thousands of drones on single CPUs and GPUs, and beyond on clusters.","excerpt":"scaling to millions of environments or thousands of swarms with thousands of drones on single CPUs and GPUs, and beyond on clusters","url":"https://arxiv.org/abs/2606.01478","relationship":"Source-stated"},{"text":"The authors report real-world experiments involving tracking without domain randomization and sampling-based obstacle avoidance at speeds exceeding half a billion steps per second.","excerpt":"Real-world experiments show Crazyflow supports both analytical-gradient-based policy learning, achieving high-accuracy tracking without domain randomization, and sampling-based obstacle avoidance at speeds exceeding half a billion steps per second.","url":"https://arxiv.org/abs/2606.01478","relationship":"Source-stated"},{"text":"The authors say they trained a recovery policy from scratch in 0.37 seconds after throwing a physical drone into the air, and caught it before it hit the ground.","excerpt":"throwing a physical drone into the air and training a recovery policy from scratch in 0.37 seconds, catching it before it hits the ground","url":"https://arxiv.org/abs/2606.01478","relationship":"Source-stated"},{"text":"The demonstration suggests a possible route to training a recovery policy during flight, but the abstract does not establish how broadly it works or how it compares with other methods.","excerpt":"Crazyflow's performance enables in-flight reinforcement learning; we demonstrate this by throwing a physical drone into the air and training a recovery policy from scratch in 0.37 seconds","url":"https://arxiv.org/abs/2606.01478","relationship":"Escape Meridian interpretation"}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2606.01478","checkedAt":"2026-11-06T01:20:06.804Z","items":[],"note":"No suitable media was found by the supported page extraction."},"assumptions":[],"openQuestion":{"attribution":"Open question","text":"How does the recovery demonstration perform across repeated trials and different flight conditions?"},"whatWouldChangeView":{"attribution":"Change condition not recorded","text":""},"editorialComparisons":[],"live":{"title":"Authors present Crazyflow, a JAX drone simulator for fast, differentiable training","summary":"The authors of Crazyflow present a drone simulator built with JAX, a software framework, that combines high-fidelity dynamics with GPU acceleration and differentiability. They say it supports model-based and data-driven methods, gradient-based and sampling-based approaches, and systems ranging from single drones to swarms.\n\nThe authors report scaling to millions of environments or thousands of swarms with thousands of drones on single CPUs and GPUs, and beyond on clusters. In real-world experiments, they report analytical-gradient-based policy learning for high-accuracy tracking without domain randomization, and sampling-based obstacle avoidance at speeds exceeding half a billion steps per second.\n\nThey also describe an in-flight reinforcement learning demonstration: a physical drone was thrown into the air, trained to recover from scratch in 0.37 seconds, and caught before hitting the ground. This suggests a possible way to train or optimize a recovery policy during flight, rather than only training before deployment; the abstract does not establish how broadly this approach works or how it compares with other methods.\n\nHow does the recovery demonstration perform across repeated trials and different flight conditions?","question":"How does the recovery demonstration perform across repeated trials and different flight conditions?","claims":[{"text":"The authors present Crazyflow as a drone simulator combining high-fidelity dynamics with JAX for speed and differentiability.","excerpt":"Crazyflow combines high-fidelity dynamics with JAX for speed and differentiability","support":"source-stated"},{"text":"The authors report that Crazyflow scales to millions of environments or thousands of swarms with thousands of drones on single CPUs and GPUs, and beyond on clusters.","excerpt":"scaling to millions of environments or thousands of swarms with thousands of drones on single CPUs and GPUs, and beyond on clusters","support":"source-stated"},{"text":"The authors report real-world experiments involving tracking without domain randomization and sampling-based obstacle avoidance at speeds exceeding half a billion steps per second.","excerpt":"Real-world experiments show Crazyflow supports both analytical-gradient-based policy learning, achieving high-accuracy tracking without domain randomization, and sampling-based obstacle avoidance at speeds exceeding half a billion steps per second.","support":"source-stated"},{"text":"The authors say they trained a recovery policy from scratch in 0.37 seconds after throwing a physical drone into the air, and caught it before it hit the ground.","excerpt":"throwing a physical drone into the air and training a recovery policy from scratch in 0.37 seconds, catching it before it hits the ground","support":"source-stated"},{"text":"The demonstration suggests a possible route to training a recovery policy during flight, but the abstract does not establish how broadly it works or how it compares with other methods.","excerpt":"Crazyflow's performance enables in-flight reinforcement learning; we demonstrate this by throwing a physical drone into the air and training a recovery policy from scratch in 0.37 seconds","support":"interpretation"}],"limitations":["The supplied abstract does not describe the methods, trial counts, flight conditions, or comparison baselines for the reported real-world experiments.","The abstract does not establish how broadly in-flight reinforcement learning works beyond the described recovery demonstration."],"mediaId":"9d5272325537e4d141361915b16baffcdc73d17d6f976a173a8dc8c495228133","correctionReason":null,"id":"92d6e8dd7a0c6465d0a96732aaa0d186db724fc30fd7cf9de2ed2459af062fbd","publicationId":"bc5a282d-44c6-4a13-89aa-57b5e86053e9","version":1,"sourceUrl":"https://arxiv.org/abs/2606.01478","sourceTitle":"Crazyflow: An Accurate, GPU-Accelerated, Differentiable Drone Simulator in JAX","publisher":"arXiv eess.SY","publishedAt":null,"originalPublished":{"value":"2026-05-31","precision":"day","basis":"publisher_citation_date"},"firstObservedAt":"2026-10-27T04:31:01.286Z","createdAt":"2026-11-06T00:17:21.759Z","historical":true,"entityIds":["arxiv"],"topicIds":["intelligence-and-discovery"],"media":[{"id":"9d5272325537e4d141361915b16baffcdc73d17d6f976a173a8dc8c495228133","kind":"article","url":"https://arxiv.org/abs/2606.01478","sourceUrl":"https://arxiv.org/abs/2606.01478","label":"Read original","provenance":"source_link","verification":{"status":"verified","checkedAt":"2026-10-30T00:31:31.033Z","method":"acquired_source","contentType":"text/html"},"recommendationReason":"Recommended by Jev for understanding this source; checked by Luna."}],"sourceMedia":{"status":"not_found","sourceUrl":"https://arxiv.org/abs/2606.01478","checkedAt":"2026-11-06T01:20:06.804Z","items":[],"note":"No suitable media was found by the supported page extraction."},"editorialDecision":{"outcome":"accept","reasonCodes":["material_development"],"rubricVersion":"stream-admission-v5","mediaId":"9d5272325537e4d141361915b16baffcdc73d17d6f976a173a8dc8c495228133","confidence":0.85,"provider":"typesafe","model":"jev-1.13.0","inputHash":"1261460e137e42ddb073b3a028c0ade98d5504107306325834e03574e44e9bb9","decidedAt":"2026-10-30T00:31:34.440Z"},"provenance":{"textHash":"aefb0240e1d6441deb627fa7b341b23aded1660b8dced653729111b29314401e","observationId":"535f178817c61c42ef7d288a62ab96a437f6d56c070d41276e67437fb76e0145","retrievedAt":"2026-10-30T00:31:30.443Z","kind":"publisher-page","preparer":{"model":"gpt-6-luna","runId":"6f7ce201-3ef6-480d-807c-4ade8da5ab5a"},"reviewer":{"model":"gpt-6-luna","runId":"ebaea321-bad0-4bd8-8d88-75840047ee9b"}},"audit":{"status":"uncertain","reviewedAt":"2026-11-06T00:17:21.759Z","model":"anthropic/claude-opus-5.5"},"sourceAttribution":{"authors":[{"name":"Schuck, Martin","url":null,"kind":"unknown"},{"name":"Rath, Marcel P.","url":null,"kind":"unknown"},{"name":"Hua, Yufei","url":null,"kind":"unknown"},{"name":"Goudar, Abhishek","url":null,"kind":"unknown"},{"name":"Zhou, SiQi","url":null,"kind":"unknown"},{"name":"Schoellig, Angela P.","url":null,"kind":"unknown"}],"authorRole":"article-author","hosts":[],"status":"stated","authorRoster":{"truncated":false,"omitted":0,"assertionGroupsOmitted":0},"participation":[{"actor":{"name":"Schuck, Martin","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Schuck, Martin","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2606.01478","locator":"head/meta[name=author|citation_author]"},"assertionId":"114659ef6fa25fa15bc8b5c82e2da38f039a60f5425ceb8a625ab41b555de4fd:participation:0","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-06T01:20:06.804Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"114659ef6fa25fa15bc8b5c82e2da38f039a60f5425ceb8a625ab41b555de4fd","sourceUrl":"https://arxiv.org/abs/2606.01478","contentSha256":"8e944ad844c11a22eac35bc943b7580d5c424ffa3cf9dd5bed746b2c9dba7325","textHash":"aefb0240e1d6441deb627fa7b341b23aded1660b8dced653729111b29314401e","observedAt":"2026-11-06T01:20:06.804Z"}},{"actor":{"name":"Rath, Marcel P.","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Rath, Marcel P.","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2606.01478","locator":"head/meta[name=author|citation_author]"},"assertionId":"114659ef6fa25fa15bc8b5c82e2da38f039a60f5425ceb8a625ab41b555de4fd:participation:1","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-06T01:20:06.804Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"114659ef6fa25fa15bc8b5c82e2da38f039a60f5425ceb8a625ab41b555de4fd","sourceUrl":"https://arxiv.org/abs/2606.01478","contentSha256":"8e944ad844c11a22eac35bc943b7580d5c424ffa3cf9dd5bed746b2c9dba7325","textHash":"aefb0240e1d6441deb627fa7b341b23aded1660b8dced653729111b29314401e","observedAt":"2026-11-06T01:20:06.804Z"}},{"actor":{"name":"Hua, Yufei","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Hua, Yufei","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2606.01478","locator":"head/meta[name=author|citation_author]"},"assertionId":"114659ef6fa25fa15bc8b5c82e2da38f039a60f5425ceb8a625ab41b555de4fd:participation:2","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-06T01:20:06.804Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"114659ef6fa25fa15bc8b5c82e2da38f039a60f5425ceb8a625ab41b555de4fd","sourceUrl":"https://arxiv.org/abs/2606.01478","contentSha256":"8e944ad844c11a22eac35bc943b7580d5c424ffa3cf9dd5bed746b2c9dba7325","textHash":"aefb0240e1d6441deb627fa7b341b23aded1660b8dced653729111b29314401e","observedAt":"2026-11-06T01:20:06.804Z"}},{"actor":{"name":"Goudar, Abhishek","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Goudar, Abhishek","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2606.01478","locator":"head/meta[name=author|citation_author]"},"assertionId":"114659ef6fa25fa15bc8b5c82e2da38f039a60f5425ceb8a625ab41b555de4fd:participation:3","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-06T01:20:06.804Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"114659ef6fa25fa15bc8b5c82e2da38f039a60f5425ceb8a625ab41b555de4fd","sourceUrl":"https://arxiv.org/abs/2606.01478","contentSha256":"8e944ad844c11a22eac35bc943b7580d5c424ffa3cf9dd5bed746b2c9dba7325","textHash":"aefb0240e1d6441deb627fa7b341b23aded1660b8dced653729111b29314401e","observedAt":"2026-11-06T01:20:06.804Z"}},{"actor":{"name":"Zhou, SiQi","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Zhou, SiQi","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2606.01478","locator":"head/meta[name=author|citation_author]"},"assertionId":"114659ef6fa25fa15bc8b5c82e2da38f039a60f5425ceb8a625ab41b555de4fd:participation:4","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-06T01:20:06.804Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"114659ef6fa25fa15bc8b5c82e2da38f039a60f5425ceb8a625ab41b555de4fd","sourceUrl":"https://arxiv.org/abs/2606.01478","contentSha256":"8e944ad844c11a22eac35bc943b7580d5c424ffa3cf9dd5bed746b2c9dba7325","textHash":"aefb0240e1d6441deb627fa7b341b23aded1660b8dced653729111b29314401e","observedAt":"2026-11-06T01:20:06.804Z"}},{"actor":{"name":"Schoellig, Angela P.","url":null,"kind":"unknown","actorId":null},"role":"author","scope":{"kind":"work-credit","locator":null},"literalCredit":"Schoellig, Angela P.","status":"stated","provenance":{"basis":"head-meta","sourceUrl":"https://arxiv.org/abs/2606.01478","locator":"head/meta[name=author|citation_author]"},"assertionId":"114659ef6fa25fa15bc8b5c82e2da38f039a60f5425ceb8a625ab41b555de4fd:participation:5","identity":{"version":"entity-identity-v1","status":"unresolved","actorId":null,"canonicalActorId":null,"candidateActorIds":[],"assertions":[],"sourceObservedAt":"2026-11-06T01:20:06.804Z","knowledgeAsOf":null},"sourceVersion":{"snapshotId":"114659ef6fa25fa15bc8b5c82e2da38f039a60f5425ceb8a625ab41b555de4fd","sourceUrl":"https://arxiv.org/abs/2606.01478","contentSha256":"8e944ad844c11a22eac35bc943b7580d5c424ffa3cf9dd5bed746b2c9dba7325","textHash":"aefb0240e1d6441deb627fa7b341b23aded1660b8dced653729111b29314401e","observedAt":"2026-11-06T01:20:06.804Z"}}],"participationCoverage":{"total":6,"returned":6,"complete":true,"fullAssertions":"source-metadata-endpoint"},"distribution":null,"publisher":null,"sourceUrl":"https://arxiv.org/abs/2606.01478","snapshotId":"114659ef6fa25fa15bc8b5c82e2da38f039a60f5425ceb8a625ab41b555de4fd","contentSha256":"8e944ad844c11a22eac35bc943b7580d5c424ffa3cf9dd5bed746b2c9dba7325","textHash":"aefb0240e1d6441deb627fa7b341b23aded1660b8dced653729111b29314401e","observedAt":"2026-11-06T01:20:06.804Z","basis":"publisher-page-metadata","sourceLinks":52,"linksTruncated":false,"metadataUrl":"/api/pipeline/stream/92d6e8dd7a0c6465d0a96732aaa0d186db724fc30fd7cf9de2ed2459af062fbd/source"}}}],"audits":[{"publicationId":"bc5a282d-44c6-4a13-89aa-57b5e86053e9","verdict":"uncertain","findings":[{"dimension":"attribution","status":"pass","explanation":"The reading consistently attributes the simulator, its capabilities and its experimental results to the paper's authors as reported claims, not established facts.","excerpt":"In this work, we propose Crazyflow, a simulator designed to push the frontier of aerial robotics algorithm development"},{"dimension":"summary","status":"uncertain","explanation":"The reading tracks the abstract, but several central terms go unexplained for a general reader. 'Differentiable' or 'differentiability' appears in the title. 'Policy', 'domain randomization', 'analytical-gradient-based policy learning', 'sampling-based' and 'reinforcement learning' are also unexplained. JAX is only called 'a software framework'. The reading also omits the source's framing that large-scale synthetic simulation data is becoming a cornerstone for robot capabilities, which is the stated reason the work matters. Smallest repair: add plain-language tooltips or glosses for these terms and include the synthetic-data framing.","excerpt":"High-quality, large-scale synthetic data from simulations is becoming a cornerstone for pushing robots' capabilities."},{"dimension":"date","status":"pass","explanation":"originalPublished 2026-05-31 matches the v1 submission date. It does not use the later v3 revision on 26 Oct 2026 or the feed date.","excerpt":"[Submitted on 31 May 2026 (v1), last revised 26 Oct 2026 (this version, v3)]"},{"dimension":"support","status":"pass","explanation":"The figures for scale, the experimental results, half a billion steps per second and the 0.37-second recovery match the abstract. The in-flight learning inference is labeled conditional and its limits are stated. The question raises an open issue rather than a leading claim, and the reading invents no beneficiaries or costs.","excerpt":"Breaking the traditional train-then-deploy paradigm, Crazyflow's performance enables in-flight reinforcement learning"},{"dimension":"media","status":"pass","explanation":"The selected media is the verified direct arXiv abstract page for this paper, which is the most direct access to the source.","excerpt":null}],"createdAt":"2026-11-06T00:17:21.759Z","model":"anthropic/claude-opus-5.5"}],"correctionUrl":"https://escapemeridian.com/corrections/"}