NuPlan: A closed-loop ML-based planning benchmark for autonomous vehicles

Caesar, Holger; Kabzan, Juraj; Tan, Kok Seang; Fong, Whye Kit; Wolff, Eric W.; Lang, Alex H.; Fletcher, Luke; Beijbom, Oscar; Omari, Sammy

doi:10.48550/arxiv.2106.11810

Cited by 12 publications

(13 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Moreover, if learned human models are conditioned on the robots' actions, they can be used to reason future human responses to robot behavior, which enables the robots to proactively shape or guide human behaviors [16,37]. Another goal of building human behavior models is to build more realistic simulators that can better access or improve the model performances in human-robot interactions without real interaction with humans [10,34]. RITA aims to accomplish this goal in autonomous driving tasks and adopts several adversarial imitation learning methods to build human behavior models.…”

Section: Human Behavior Modelingmentioning

confidence: 99%

RITA: Boost Autonomous Driving Simulators with Realistic Interactive Traffic Flow

Zhu¹,

Zhang²,

Zhuang³

et al. 2022

Preprint

View full text Add to dashboard Cite

High-quality traffic flow generation is the core module in building simulators for autonomous driving. However, the majority of available simulators are incapable of replicating traffic patterns that accurately reflect the various features of real-world data while also simulating human-like reactive responses to the tested autopilot driving strategies. Taking one step forward to addressing such a problem, we propose Realistic Interactive TrAffic flow (RITA) as an integrated component of existing driving simulators to provide high-quality traffic flow for the evaluation and optimization of the tested driving strategies. RITA is developed with fidelity, diversity, and controllability in consideration, and consists of two core modules called RITABackend and RITAKit. RITABackend is built to support vehicle-wise control and provide traffic generation models from real-world datasets, while RITAKit is developed with easy-to-use interfaces for controllable traffic generation via RITABackend. We demonstrate RITA's capacity to create diversified and high-fidelity traffic simulations in several highly interactive highway scenarios. The experimental findings demonstrate that our produced RITA traffic flows meet all three design goals, hence enhancing the completeness of driving strategy evaluation. Moreover, we showcase the possibility for further improvement of baseline strategies through online fine-tuning with RITA traffic flows.

show abstract

Section: Human Behavior Modelingmentioning

confidence: 99%

RITA: Boost Autonomous Driving Simulators with Realistic Interactive Traffic Flow

Zhu¹,

Zhang²,

Zhuang³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…We conduct closed-loop simulation in nuPlan [41] to evaluate the proposed approach. The closed-loop planner consists of three modules, a trajectory predictor that generates the unconditioned trajectory prediction, a route planner that distills lane information and reference trajectory from the lane graph, and IJP that plans the trajectory, as shown in Fig.…”

Section: A Simulation Evaluation Setupmentioning

confidence: 99%

Motion Planning for Autonomous Vehicles Based on Sequential Optimization

Diachuk¹,

Easa²

2022

Preprint

View full text Add to dashboard Cite

This study presents the substantiation, development, and analysis of a technique for planning the autonomous vehicle (AV) motion reference parameters. The trajectory plan, speed and acceleration distributions, including other AV's kinematic parameters, are determined using sequential optimization. The study objectives are based on an analysis of the fundamental problems of AV motion planning summarized in this area's latest publications. The proposed approach combines the basic principles of the finite element method (FEM) and nonlinear optimization with nonlinear constraints. First, the generalization on representing an investigated function by finite elements (FE) is briefly described. A one-dimension FE with two nodes and three degrees of freedom (DOF) in a node was chosen as the basic one, corresponding to the 5th-degree polynomial. Next, a method for determining the motion trajectory is presented. The following are considered: formation of a restricted space for the AV's allowable maneuvering, the geometry of motion trajectory and its relation with vehicle steerability parameters, cost functions and their influences on the desirable trajectory's nature, compliance of nonlinear restrictions of the node parameters with the motion area boundaries. At the second stage, a technique for optimizing AV speed and acceleration redistribution is presented. The model considers possible combinations of cost functions, conditions of limiting the kinematic parameters with the tire slip critical speed, maximum speed level, maximum longitudinal acceleration, and critical lateral acceleration. In the simulation section, several variants of trajectories were searched and compared. Several versions of distributing the longitudinal speed and acceleration curves are determined, and their comparative analysis is fulfilled. At the end of the paper, the advantages and drawbacks of the proposed technique are noted. The conclusion is made regarding the options for improving the method in further studies.

show abstract

“…Instead, we analyze the limitations of existing models and propose simple yet effective strategies to estimate occluded positions and improve temporal coherence. In addition, our streaming forecasting is broadly relevant to open-loop planning [3] which similarly models a streaming world, with the key difference of predicting the motion of the ego-vehicle, while we focus on surrounding agents.…”

Section: A Motion Forecasting In Autonomous Drivingmentioning

confidence: 99%

Learning hierarchical behavior and motion planning for autonomous driving

Wang

Zhang

et al. 2020

2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)

View full text Add to dashboard Cite

Trajectory forecasting is a widely-studied problem for autonomous navigation. However, existing benchmarks evaluate forecasting based on independent snapshots of trajectories, which are not representative of real-world applications that operate on a continuous stream of data. To bridge this gap, we introduce a benchmark that continuously queries future trajectories on streaming data and we refer to it as "streaming forecasting." Our benchmark inherently captures the disappearance and re-appearance of agents, presenting the emergent challenge of forecasting for occluded agents, which is a safetycritical problem yet overlooked by snapshot-based benchmarks. Moreover, forecasting in the context of continuous timestamps naturally asks for temporal coherence between predictions from adjacent timestamps. Based on this benchmark, we further provide solutions and analysis for streaming forecasting. We propose a plug-and-play meta-algorithm called "Predictive Streamer" that can adapt any snapshot-based forecaster into a streaming forecaster. Our algorithm estimates the states of occluded agents by propagating their positions with multimodal trajectories, and leverages differentiable filters to ensure temporal consistency. Both occlusion reasoning and temporal coherence strategies significantly improve forecasting quality, resulting in 25% smaller endpoint errors for occluded agents and 10-20% smaller fluctuations of trajectories. Our work is intended to generate interest within the community by highlighting the importance of addressing motion forecasting in its intrinsic streaming setting. Code is available at https://github.com/ziqipang/StreamingForecasting.

show abstract

NuPlan: A closed-loop ML-based planning benchmark for autonomous vehicles

Cited by 12 publications

References 18 publications

RITA: Boost Autonomous Driving Simulators with Realistic Interactive Traffic Flow

RITA: Boost Autonomous Driving Simulators with Realistic Interactive Traffic Flow

Motion Planning for Autonomous Vehicles Based on Sequential Optimization

Learning hierarchical behavior and motion planning for autonomous driving

Contact Info

Product

Resources

About