Consistent Depth Estimation in Data-Driven Simulation for Autonomous Driving
In this work we propose consistent depth estimation for viewpoint reconstruction in data-driven simulation, combining aspects of learning-based monocular depth prediction and structure-from-motion to increase temporal video depth accuracy. We demonstrate efficacy in VISTA, an end-to-end autonomous v...
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Published: |
Massachusetts Institute of Technology
2022
|
| Online Access: | https://hdl.handle.net/1721.1/139136 |
| _version_ | 1811093169332289536 |
|---|---|
| author | Beveridge, Matthew |
| author2 | Rus, Daniela |
| author_facet | Rus, Daniela Beveridge, Matthew |
| author_sort | Beveridge, Matthew |
| collection | MIT |
| description | In this work we propose consistent depth estimation for viewpoint reconstruction in data-driven simulation, combining aspects of learning-based monocular depth prediction and structure-from-motion to increase temporal video depth accuracy. We demonstrate efficacy in VISTA, an end-to-end autonomous vehicle simulation engine capable of training robust control policies directly applicable to the real-world. Taking advantage of geometrically consistent depth map estimations, we see a several order of magnitude improvement in whole-frame depth accuracy averaged over the course of input traces compared to VISTA’s current depth method, and a 39% reduction in intra-frame depth variance compared to current state of the art methods (i.e. Monodepth2) while maintaining similar error. Better depth enables more accurate viewpoint reconstruction thus improving the training of reinforcement learning (RL) control policies in simulation, increasing RL-based control’s practicality. We train several end-to-end policy gradient models in varying versions of VISTA, each utilizing a different depth method, and see that end-to-end models trained in the consistent depth version of VISTA deviate least from the human driven center line. |
| first_indexed | 2024-09-23T15:40:26Z |
| format | Thesis |
| id | mit-1721.1/139136 |
| institution | Massachusetts Institute of Technology |
| last_indexed | 2024-09-23T15:40:26Z |
| publishDate | 2022 |
| publisher | Massachusetts Institute of Technology |
| record_format | dspace |
| spelling | mit-1721.1/1391362022-01-15T03:08:26Z Consistent Depth Estimation in Data-Driven Simulation for Autonomous Driving Beveridge, Matthew Rus, Daniela Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science In this work we propose consistent depth estimation for viewpoint reconstruction in data-driven simulation, combining aspects of learning-based monocular depth prediction and structure-from-motion to increase temporal video depth accuracy. We demonstrate efficacy in VISTA, an end-to-end autonomous vehicle simulation engine capable of training robust control policies directly applicable to the real-world. Taking advantage of geometrically consistent depth map estimations, we see a several order of magnitude improvement in whole-frame depth accuracy averaged over the course of input traces compared to VISTA’s current depth method, and a 39% reduction in intra-frame depth variance compared to current state of the art methods (i.e. Monodepth2) while maintaining similar error. Better depth enables more accurate viewpoint reconstruction thus improving the training of reinforcement learning (RL) control policies in simulation, increasing RL-based control’s practicality. We train several end-to-end policy gradient models in varying versions of VISTA, each utilizing a different depth method, and see that end-to-end models trained in the consistent depth version of VISTA deviate least from the human driven center line. M.Eng. 2022-01-14T14:52:03Z 2022-01-14T14:52:03Z 2021-06 2021-06-17T20:12:51.932Z Thesis https://hdl.handle.net/1721.1/139136 In Copyright - Educational Use Permitted Copyright MIT http://rightsstatements.org/page/InC-EDU/1.0/ application/pdf Massachusetts Institute of Technology |
| spellingShingle | Beveridge, Matthew Consistent Depth Estimation in Data-Driven Simulation for Autonomous Driving |
| title | Consistent Depth Estimation in Data-Driven Simulation for Autonomous Driving |
| title_full | Consistent Depth Estimation in Data-Driven Simulation for Autonomous Driving |
| title_fullStr | Consistent Depth Estimation in Data-Driven Simulation for Autonomous Driving |
| title_full_unstemmed | Consistent Depth Estimation in Data-Driven Simulation for Autonomous Driving |
| title_short | Consistent Depth Estimation in Data-Driven Simulation for Autonomous Driving |
| title_sort | consistent depth estimation in data driven simulation for autonomous driving |
| url | https://hdl.handle.net/1721.1/139136 |
| work_keys_str_mv | AT beveridgematthew consistentdepthestimationindatadrivensimulationforautonomousdriving |