Learn-to-Race: A Multimodal Control Environment for Autonomous Racing

Herman, James; Francis, Jonathan; Ganju, Siddha; Chen, Bingqing; Koul, Anirudh; Gupta, Abhinav; Skabelkin, Alexey; Жуков, И. А.; Kumskoy, Max; Nyberg, Eric

doi:10.1109/iccv48922.2021.00965

Cited by 17 publications

(20 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The maps in these traffic simulators are either created by humans or based on real-world data. For example, Learnto-Race (L2R) [169] platform has three racetracks in its racing simulator, all of which are based on real-world racetracks. Human-designed maps can be further classified into two types: rule-based maps and procedurally generated maps.…”

Section: Map Sourcementioning

confidence: 99%

A Survey on Safety-Critical Driving Scenario Generation -- A Methodological Perspective

Ding¹,

Xu²,

Arief³

et al. 2022

Preprint

View full text Add to dashboard Cite

Autonomous driving systems have witnessed a significant development during the past years thanks to the advance in machine learning-enabled sensing and decision-making algorithms. One critical challenge for their massive deployment in the real world is their safety evaluation. Most existing driving systems are still trained and evaluated on naturalistic scenarios collected from daily life or heuristically-generated adversarial ones. However, the large population of cars, in general, leads to an extremely low collision rate, indicating that the safety-critical scenarios are rare in the collected real-world data. Thus, methods to artificially generate scenarios become crucial to measure the risk and reduce the cost. In this survey, we focus on the algorithms of safety-critical scenario generation in autonomous driving. We first provide a comprehensive taxonomy of existing algorithms by dividing them into three categories: data-driven generation, adversarial generation, and knowledge-based generation. Then, we discuss useful tools for scenario generation, including simulation platforms and packages. Finally, we extend our discussion to five main challenges of current works -fidelity, efficiency, diversity, transferability, controllability -and research opportunities lighted up by these challenges.

show abstract

Section: Map Sourcementioning

confidence: 99%

A Survey on Safety-Critical Driving Scenario Generation -- A Methodological Perspective

Ding¹,

Xu²,

Arief³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Learn-to-Race 3 [20,10] is a recent Gym-compliant open-source framework based on a high-fidelity racing simulator developed by Arrival, able to capture complex vehicle dynamics and to render 3D photorealistic views.…”

Section: Learn-to-racementioning

confidence: 99%

MicroRacer: a didactic environment for Deep Reinforcement Learning

Asperti¹,

Brutto²

2022

Preprint

View full text Add to dashboard Cite

MicroRacer is a simple, open source environment inspired by car racing especially meant for the didactics of Deep Reinforcement Learning. The complexity of the environment has been explicitly calibrated to allow users to experiment with many different methods, networks and hyperparameters settings without requiring sophisticated software or the need of exceedingly long training times. Baseline agents for major learning algorithms such as DDPG, PPO, SAC, TD2 and DSAC are provided too, along with a preliminary comparison in terms of training time and performance.

show abstract

“…In our prior work (Herman et al, 2021), we released Learn-to-Race (L2R): a python-and pytorchbased open-source, OpenAI Gym-compliant (Brockman et al, 2016) framework which leverages a high-fidelity racing simulator or real-world vehicle interface (e.g., implemented in the Robot Operating System; ROS (Quigley et al, 2009)), for training and deployment of machine learning algorithms for continuous control contexts. We utilise the Arrival Ltd. autonomous racing simulator, which not only captures complex vehicle dynamics and renders photo-realistic views, but also plays a key role in bringing autonomous racing technology to real life-through official contribution to the Roborace Challenge series, the world's first extreme competition of teams developing autonomous Learn-to-Race environment.…”

Section: Learn-to-race Frameworkmentioning

confidence: 99%

“…L2R supports the use of an autonomous driving simulator, as a backend API service, for developing policy algorithms for vehicle control. Currently, L2R uses the Arrival simulator (Herman et al, 2021), which is a powerful tool for the development and testing of autonomous vehicles. It is based on Unreal Engine 4 and includes such features as: (i) a vehicle prototyping framework; (ii) full software-in-the-loop (SIL) simulation, to model all vehicle control devices; (iii) controller area network (CAN) bus interface; (iv) camera, inertial measurement unit (IMU), light detection and ranging (LiDAR), ultrasonic, and radar sensor models; (v) semantic segmentation; (vi) sensor placement and configuration facilities; (vii) V2V/V2I interface subsystem; (vii) dynamic racing scenario creation; (viii) race track generation from scanned datasets; (ix) support for full integration with the CARLA simulator (Dosovitskiy et al, 2017b); and (x) an application programming interface (API), which is automatically generated based on C++ code analysis.…”

Section: Learn-to-race Frameworkmentioning

confidence: 99%

“…Autonomous agents are required to make sub-second, safety-critical decisions, in a complex, fast-changing environment; both perception and control must remain robust to distribution shifts, novel road features, and unseen obstacles, in order to facilitate cross-domain safety and performance. In Herman et al (2021), we released an open-source, high-fidelity simulation environment for high-speed racing, Learn-to-Race 1 (L2R), with the hope to democratize autonomous racing research. In order to encourage use of L2R and to facilitate standardized benchmarking, we launch the Learn-to-Race autonomous racing virtual challenge, with multi-institutional support from Carnegie Mellon University, Arrival Ltd., AICrowd, Amazon Web Services, and Honda Research.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Learn-to-Race Challenge 2022: Benchmarking Safe Learning and Cross-domain Generalisation in Autonomous Racing

Francis¹,

Chen²,

Ganju³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

We present the results of our autonomous racing virtual challenge, based on the newly-released Learn-to-Race (L2R) simulation framework, which seeks to encourage interdisciplinary research in autonomous driving and to help advance the state of the art on a realistic benchmark. Analogous to racing being used to test cutting-edge vehicles, we envision autonomous racing to serve as a particularly challenging proving ground for autonomous agents as: (i) they need to make sub-second, safety-critical decisions in a complex, fast-changing environment; and (ii) both perception and control must be robust to distribution shifts, novel road features, and unseen obstacles. Thus, the main goal of the challenge is to evaluate the joint safety, performance, and generalisation capabilities of reinforcement learning agents on multi-modal perception, through a two-stage process. In the first stage of the challenge, we evaluate an autonomous agent's ability to drive as fast as possible, while adhering to safety constraints. In the second stage, we additionally require the agent to adapt to an unseen racetrack through safe exploration. In this paper, we describe the new L2R Task 2.0 benchmark, with refined metrics and baseline approaches. We also provide an overview of deployment, evaluation, and rankings for the inaugural instance of the L2R Autonomous Racing Virtual Challenge (supported by Carnegie Mellon University, Arrival Ltd., AICrowd, Amazon Web Services, and Honda Research), which officially used the new L2R Task 2.0 benchmark and received over 20,100 views, 437 active participants, 46 teams, and 733 model submissions-from 88+ unique institutions, in 58+ different countries. Finally, we release leaderboard results from the challenge and provide description of the two top-ranking approaches in cross-domain model transfer, across multiple sensor configurations and simulated races.

show abstract

Learn-to-Race: A Multimodal Control Environment for Autonomous Racing

Cited by 17 publications

References 11 publications

A Survey on Safety-Critical Driving Scenario Generation -- A Methodological Perspective

A Survey on Safety-Critical Driving Scenario Generation -- A Methodological Perspective

MicroRacer: a didactic environment for Deep Reinforcement Learning

Learn-to-Race Challenge 2022: Benchmarking Safe Learning and Cross-domain Generalisation in Autonomous Racing

Contact Info

Product

Resources

About