{"paper":{"title":"CycleRL: Sim-to-Real Deep Reinforcement Learning for Robust Autonomous Bicycle Control","license":"http://creativecommons.org/licenses/by/4.0/","headline":"CycleRL trains a PPO policy in simulation that transfers directly to physical bicycle hardware for balance and tracking.","cross_cats":[],"primary_cat":"cs.RO","authors_text":"Gelu Liu, Junliang Wu, Songyuan Li, Teng Wang, Xiangwei Zhu, Zhijie Wu","submitted_at":"2026-03-16T09:17:51Z","abstract_excerpt":"Autonomous bicycles offer a promising agile solution for urban mobility and last-mile logistics. However, conventional control strategies often struggle with underactuated nonlinear dynamics, suffering from sensitivity to model mismatches and limited adaptability to real-world uncertainties. To address this, we develop CycleRL, a comprehensive sim-to-real framework for robust autonomous bicycle control. Our approach establishes a direct perception-to-action mapping within the high-fidelity NVIDIA Isaac Sim environment, leveraging Proximal Policy Optimization (PPO) to optimize the control polic"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"CycleRL achieves a 99.90% balance success rate, 1.15° heading tracking error, and 0.18 m/s velocity tracking error in simulation, with successful hardware deployment that validates DRL as offering superior adaptability over traditional methods.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That systematic domain randomization over a limited set of simulation parameters is sufficient to cover all real-world uncertainties and enable zero-shot transfer to physical hardware without additional adaptation or fine-tuning.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"CycleRL trains a PPO policy in Isaac Sim with domain randomization to achieve 99.9% balance success and direct hardware transfer for autonomous bicycle control.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"CycleRL trains a PPO policy in simulation that transfers directly to physical bicycle hardware for balance and tracking.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"a970ae49f726748f34243f01a428fefa1650167ea5d9ee3a7ae23cff98d93ace"},"source":{"id":"2603.15013","kind":"arxiv","version":3},"verdict":{"id":"ec36a41f-82c1-4726-b4a6-caf623a7545f","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T10:41:17.457744Z","strongest_claim":"CycleRL achieves a 99.90% balance success rate, 1.15° heading tracking error, and 0.18 m/s velocity tracking error in simulation, with successful hardware deployment that validates DRL as offering superior adaptability over traditional methods.","one_line_summary":"CycleRL trains a PPO policy in Isaac Sim with domain randomization to achieve 99.9% balance success and direct hardware transfer for autonomous bicycle control.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That systematic domain randomization over a limited set of simulation parameters is sufficient to cover all real-world uncertainties and enable zero-shot transfer to physical hardware without additional adaptation or fine-tuning.","pith_extraction_headline":"CycleRL trains a PPO policy in simulation that transfers directly to physical bicycle hardware for balance and tracking."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2603.15013/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"5f9cd9c3fd499629b3a9f3042127cedd64ceac46f0af372e23ad2d9aa594d7cc"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}