Integrating Modular Pipelines with End-to-End Learning: A Hybrid Approach for Robust and Reliable Autonomous Driving Systems
Autonomous driving navigation relies on diverse approaches, each with advantages and limitations depending on various factors. For HD maps, modular systems excel, while end-to-end methods dominate mapless scenarios. However, few leverage the strengths of both. This paper innovates by proposing a hyb...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2024-03-01
|
| Series: | Sensors |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1424-8220/24/7/2097 |
| _version_ | 1797212053150105600 |
|---|---|
| author | Luis Alberto Rosero Iago Pachêco Gomes Júnior Anderson Rodrigues da Silva Carlos André Przewodowski Denis Fernando Wolf Fernando Santos Osório |
| author_facet | Luis Alberto Rosero Iago Pachêco Gomes Júnior Anderson Rodrigues da Silva Carlos André Przewodowski Denis Fernando Wolf Fernando Santos Osório |
| author_sort | Luis Alberto Rosero |
| collection | DOAJ |
| description | Autonomous driving navigation relies on diverse approaches, each with advantages and limitations depending on various factors. For HD maps, modular systems excel, while end-to-end methods dominate mapless scenarios. However, few leverage the strengths of both. This paper innovates by proposing a hybrid architecture that seamlessly integrates modular perception and control modules with data-driven path planning. This innovative design leverages the strengths of both approaches, enabling a clear understanding and debugging of individual components while simultaneously harnessing the learning power of end-to-end approaches. Our proposed architecture achieved first and second place in the 2023 CARLA Autonomous Driving Challenge’s SENSORS and MAP tracks, respectively. These results demonstrate the architecture’s effectiveness in both map-based and mapless navigation. We achieved a driving score of <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>41.56</mn></mrow></semantics></math></inline-formula> and the highest route completion of <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>86.03</mn></mrow></semantics></math></inline-formula> in the MAP track of the CARLA Challenge leaderboard 1, and driving scores of <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>35.36</mn></mrow></semantics></math></inline-formula> and <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>1.23</mn></mrow></semantics></math></inline-formula> in the CARLA Challenge SENSOR track with route completions of <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>85.01</mn></mrow></semantics></math></inline-formula> and <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>9.55</mn></mrow></semantics></math></inline-formula>, for, respectively, leaderboard 1 and 2. The results of leaderboard 2 raised the hybrid architecture to the first position, winning the edition of the 2023 CARLA Autonomous Driving Competition. |
| first_indexed | 2024-04-24T10:36:16Z |
| format | Article |
| id | doaj.art-f37c4901884b4423a41458dfaefe0e1e |
| institution | Directory Open Access Journal |
| issn | 1424-8220 |
| language | English |
| last_indexed | 2024-04-24T10:36:16Z |
| publishDate | 2024-03-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Sensors |
| spelling | doaj.art-f37c4901884b4423a41458dfaefe0e1e2024-04-12T13:26:12ZengMDPI AGSensors1424-82202024-03-01247209710.3390/s24072097Integrating Modular Pipelines with End-to-End Learning: A Hybrid Approach for Robust and Reliable Autonomous Driving SystemsLuis Alberto Rosero0Iago Pachêco Gomes1Júnior Anderson Rodrigues da Silva2Carlos André Przewodowski3Denis Fernando Wolf4Fernando Santos Osório5Institute of Mathematics and Computer Science, University of São Paulo, Ave. Trabalhador São-Carlense, 400, São Carlos 13564-002, SP, BrazilInstitute of Mathematics and Computer Science, University of São Paulo, Ave. Trabalhador São-Carlense, 400, São Carlos 13564-002, SP, BrazilInstitute of Mathematics and Computer Science, University of São Paulo, Ave. Trabalhador São-Carlense, 400, São Carlos 13564-002, SP, BrazilInstitute of Mathematics and Computer Science, University of São Paulo, Ave. Trabalhador São-Carlense, 400, São Carlos 13564-002, SP, BrazilInstitute of Mathematics and Computer Science, University of São Paulo, Ave. Trabalhador São-Carlense, 400, São Carlos 13564-002, SP, BrazilInstitute of Mathematics and Computer Science, University of São Paulo, Ave. Trabalhador São-Carlense, 400, São Carlos 13564-002, SP, BrazilAutonomous driving navigation relies on diverse approaches, each with advantages and limitations depending on various factors. For HD maps, modular systems excel, while end-to-end methods dominate mapless scenarios. However, few leverage the strengths of both. This paper innovates by proposing a hybrid architecture that seamlessly integrates modular perception and control modules with data-driven path planning. This innovative design leverages the strengths of both approaches, enabling a clear understanding and debugging of individual components while simultaneously harnessing the learning power of end-to-end approaches. Our proposed architecture achieved first and second place in the 2023 CARLA Autonomous Driving Challenge’s SENSORS and MAP tracks, respectively. These results demonstrate the architecture’s effectiveness in both map-based and mapless navigation. We achieved a driving score of <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>41.56</mn></mrow></semantics></math></inline-formula> and the highest route completion of <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>86.03</mn></mrow></semantics></math></inline-formula> in the MAP track of the CARLA Challenge leaderboard 1, and driving scores of <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>35.36</mn></mrow></semantics></math></inline-formula> and <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>1.23</mn></mrow></semantics></math></inline-formula> in the CARLA Challenge SENSOR track with route completions of <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>85.01</mn></mrow></semantics></math></inline-formula> and <inline-formula><math xmlns="http://www.w3.org/1998/Math/MathML" display="inline"><semantics><mrow><mn>9.55</mn></mrow></semantics></math></inline-formula>, for, respectively, leaderboard 1 and 2. The results of leaderboard 2 raised the hybrid architecture to the first position, winning the edition of the 2023 CARLA Autonomous Driving Competition.https://www.mdpi.com/1424-8220/24/7/2097autonomous drivinghybrid architecturemodularend-to-endpath planningCARLA simulator |
| spellingShingle | Luis Alberto Rosero Iago Pachêco Gomes Júnior Anderson Rodrigues da Silva Carlos André Przewodowski Denis Fernando Wolf Fernando Santos Osório Integrating Modular Pipelines with End-to-End Learning: A Hybrid Approach for Robust and Reliable Autonomous Driving Systems Sensors autonomous driving hybrid architecture modular end-to-end path planning CARLA simulator |
| title | Integrating Modular Pipelines with End-to-End Learning: A Hybrid Approach for Robust and Reliable Autonomous Driving Systems |
| title_full | Integrating Modular Pipelines with End-to-End Learning: A Hybrid Approach for Robust and Reliable Autonomous Driving Systems |
| title_fullStr | Integrating Modular Pipelines with End-to-End Learning: A Hybrid Approach for Robust and Reliable Autonomous Driving Systems |
| title_full_unstemmed | Integrating Modular Pipelines with End-to-End Learning: A Hybrid Approach for Robust and Reliable Autonomous Driving Systems |
| title_short | Integrating Modular Pipelines with End-to-End Learning: A Hybrid Approach for Robust and Reliable Autonomous Driving Systems |
| title_sort | integrating modular pipelines with end to end learning a hybrid approach for robust and reliable autonomous driving systems |
| topic | autonomous driving hybrid architecture modular end-to-end path planning CARLA simulator |
| url | https://www.mdpi.com/1424-8220/24/7/2097 |
| work_keys_str_mv | AT luisalbertorosero integratingmodularpipelineswithendtoendlearningahybridapproachforrobustandreliableautonomousdrivingsystems AT iagopachecogomes integratingmodularpipelineswithendtoendlearningahybridapproachforrobustandreliableautonomousdrivingsystems AT juniorandersonrodriguesdasilva integratingmodularpipelineswithendtoendlearningahybridapproachforrobustandreliableautonomousdrivingsystems AT carlosandreprzewodowski integratingmodularpipelineswithendtoendlearningahybridapproachforrobustandreliableautonomousdrivingsystems AT denisfernandowolf integratingmodularpipelineswithendtoendlearningahybridapproachforrobustandreliableautonomousdrivingsystems AT fernandosantososorio integratingmodularpipelineswithendtoendlearningahybridapproachforrobustandreliableautonomousdrivingsystems |