Trajectory prediction of traffic agents at urban intersections through learned interactions

Sarkar, Atrisha; Czarnecki, Krzysztof; Angus, Matt; Li, Changjian; Waslander, Steven L.

doi:10.1109/itsc.2017.8317731

Cited by 21 publications

(13 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Tightly integrated planning and control algorithms have been created using a combination of machine learning and model-based techniques [3,4]. These algorithms exploit the enriched sensor and environment data to enable the AV to operate e ciently and safely throughout their operating envelope [5]. The researchers are establishing minimal hierarchical safety requirements for autonomous functions, delity-aware simulations that can rapidly evaluate large numbers of challenging driving scenarios, and safety assurance methods for systems that rely on machine learning [6,7].…”

Section: Research and Educational Directionsmentioning

confidence: 99%

Research and Educational Programs for Connected and Autonomous Vehicles at the University of Waterloo

Mckenzie

McPhee

2017

Mechanical Engineering

View full text Add to dashboard Cite

This article presents an overview of the research and educational programs for connected and autonomous vehicles at the University of Waterloo (UWaterloo). UWaterloo is Canada’s largest engineering school, with 9,500 engineering students and 309 engineering faculty. The University of Waterloo Centre for Automotive Research (WatCAR) for faculty, staff and students is contributing to the development of in-vehicle systems education programs for connected and autonomous vehicles (CAVs) at Waterloo. Over 130 Waterloo faculty, 110 from engineering, are engaged in WatCAR’s automotive and transportation systems research programs. The school’s CAV efforts leverage WatCAR research expertise from five areas: (1) Connected and Autonomous; (2) Software and Data; (3) Lightweighting and Fabrication; (4) Structure and Safety; and (5) Advanced Powertrain and Emissions. Foundational and operational artificial intelligence expertise from the University of Waterloo Artificial Intelligence Institute complements the autonomous driving efforts, in disciplines that include neural networks, pattern analysis and machine learning.

show abstract

Section: Research and Educational Directionsmentioning

confidence: 99%

Research and Educational Programs for Connected and Autonomous Vehicles at the University of Waterloo

Mckenzie

McPhee

2017

Mechanical Engineering

View full text Add to dashboard Cite

show abstract

“…Similar approaches entailed training an SVM and random decision forests [21], and an artificial neural network was applied to design a motion predictor. The influence network, which simultaneously considers agents and the environment, was trained and evaluated by driving data extracted from a surveillance camera at an intersection [33]. Investigators have also used Long Short-Term Memory (LSTM)-RNN to predict the future intention of targets at a roundabout [34].…”

Section: Introductionmentioning

confidence: 99%

Surround Vehicle Motion Prediction Using LSTM-RNN for Motion Planning of Autonomous Vehicles at Multi-Lane Turn Intersections

Jeong

Kim

2020

IEEE Open J. Intell. Transp. Syst.

View full text Add to dashboard Cite

This paper presents a surround vehicle motion prediction algorithm for multi-lane turn intersections using a Long Short-Term Memory (LSTM)-based Recurrent Neural Network (RNN). The motion predictor is trained using the states of subject and surrounding vehicles, which are collected by sensors mounted on an autonomous vehicle. Data on 484 vehicle trajectories were collected from real traffic situations at multi-lane turn intersections. 11,662 and 4,998 samples acquired from the vehicle trajectories were used to train and evaluate the networks, respectively. A motion planner based on Model Predictive Control (MPC) is designed to determine the longitudinal acceleration command based on the predicted states of surrounding vehicles. The future states of the subject vehicle derived by MPC is used as an input feature to reflect the interaction of subject and target vehicles in LSTM-RNN based motion predictor. The proposed algorithm was evaluated in terms of its accuracy and its effects on the motion planning algorithm based on the driving data sets. The improved prediction accuracy substantially increased safety by bounding the prediction error within the safety margin. The application results of the proposed predictor demonstrate the improved recognition timing of the preceding vehicle and the similarity of longitudinal acceleration with drivers. INDEX TERMS Autonomous vehicle, intersection driving data, motion prediction, machine learning, recurrent neural network, long short-term memory, model predictive control.

show abstract

“…For the behavior planner (which is the component in ADS responsible for tactical and high-level decision making), this means that simulation environments should be able to model behavior of other traffic users in a way that is reflective of the real-world behavior. Popular approaches design this behavior in several ways: expert-driven, where designers program the motion and behavior of the users [2], data-driven where a model of behavior is learnt from observations and naturalistic driving datasets [3], or a hybrid model that uses a combination of both [4] [5]. Although it is possible to design models that learn from real-world data, a major challenge in any approach is the generation of unusual or atypical behavior that is not readily observed in the data, such as crashes or near-miss scenarios.…”

Section: Introductionmentioning

confidence: 99%

A behavior driven approach for sampling rare event situations for autonomous vehicles

Sarkar

Czamecki

2019

2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)

View full text Add to dashboard Cite

Performance evaluation of urban autonomous vehicles requires a realistic model of the behavior of other road users in the environment. Learning such models from data involves collecting naturalistic data of real-world human behavior. In many cases, acquisition of this data can be prohibitively expensive or intrusive. Additionally, the available data often contain only typical behaviors and exclude behaviors that are classified as rare events. To evaluate the performance of AV in such situations, we develop a model of traffic behavior based on the theory of bounded rationality. Based on the experiments performed on a large naturalistic driving data, we show that the developed model can be applied to estimate probability of rare events, as well as to generate new traffic situations.

show abstract

Trajectory prediction of traffic agents at urban intersections through learned interactions

Cited by 21 publications

References 21 publications

Research and Educational Programs for Connected and Autonomous Vehicles at the University of Waterloo

Research and Educational Programs for Connected and Autonomous Vehicles at the University of Waterloo

Surround Vehicle Motion Prediction Using LSTM-RNN for Motion Planning of Autonomous Vehicles at Multi-Lane Turn Intersections

A behavior driven approach for sampling rare event situations for autonomous vehicles

Contact Info

Product

Resources

About