Research map / Reinforcement learning / Visual and representation RL
Reinforcement Learning Meets Visual Odometry
Nico Messikommer, Giovanni Cioffi, Mathias Gehrig, Davide Scaramuzza
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add an RL agent for keyframe and grid-size selection to DSO | DSO · EuRoC | ATE avg (m): 0.594 -> 0.483 | Improved |
| add an RL agent to SVO | SVO · EuRoC V103 | tracking success: SVO fails -> 0.394 | Improved |
| add an RL agent to SVO | SVO · TUM-RGBD | ATE avg (m): 0.471 -> 0.422 | Improved |
| add an RL agent to DSO on TUM-RGBD (rolling shutter) | DSO · TUM-RGBD | sequences completed: two more sequences completed; some per-sequence errors higher | Mixed |
| add an RL agent to ORB-SLAM3 | ORB-SLAM3 · EuRoC | ATE avg (m): 0.573 -> 0.475 | Improved |
| add keyframe selection action to the agent (ablation) | RL SVO · EuRoC | robustness | Improved |
| add grid size selection action (ablation) | RL SVO · EuRoC | ATE avg (m): 0.799 -> 0.696 | Improved |
| add keypoint inputs via Variable Encoder (ablation) | RL SVO · EuRoC | sequences completed | Improved |
| train with privileged critic (ablation) | RL SVO · EuRoC | ATE avg (m): 0.754 -> 0.696 | Improved |
| use RL agent for keyframe selection | SVO · EuRoC | runtime (ms): 9.06 -> 7.40 (SVO); 34.45 -> 32.29 (DSO) | Improved |
| RL-tuned classical VO versus learned hybrid VO (DPVO) | RL DSO / RL SVO · EuRoC | ATE avg (m): DPVO 0.105 vs RL DSO 0.320 vs RL SVO 0.972 | Hurt |
Closest papers by meaning
- Dual-Agent Reinforcement Learning for Adaptive and Cost-Aware Visual-Inertial Odometry (CVPR 2026 · similarity 0.93)
- Modality-Invariant Visual Odometry for Embodied Vision (CVPR 2023 · similarity 0.89)
- Seeing through Uncertainty: Robust Task-Oriented Optimization in Visual Navigation (NeurIPS 2025 · similarity 0.87)
- COVR: Collaborative Optimization of VLMs and RL Agent for Visual-Based Control (AAAI 2026 · similarity 0.86)
- LEAP-VO: Long-term Effective Any Point Tracking for Visual Odometry (CVPR 2024 · similarity 0.86)
- HASARD: A Benchmark for Vision-Based Safe Reinforcement Learning in Embodied Agents (ICLR 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
