Multi-Modal Trajectory Prediction of Surrounding Vehicles with Maneuver based LSTMs

By Nachiket Deo, Mohan Manubhai Trivedi. 2018 IEEE Intelligent Vehicles Symposium (IV), 2018

475 citations recorded by the source corpus.

To safely and efficiently navigate through complex traffic scenarios, autonomous vehicles need to have the ability to predict the future motion of surrounding vehicles. Multiple interacting agents, the multi-modal nature of driver behavior, and the inherent uncertainty involved in the task make motion prediction of surrounding vehicles a challenging problem. In this paper, we present an LSTM model for interaction aware motion prediction of surrounding vehicles on freeways. Our model assigns confidence values to maneuvers being performed by vehicles and outputs a multi-modal distribution over future motion based on them. We compare our approach with the prior art for vehicle motion prediction on the publicly available NGSIM US-101 and I-80 datasets. Our results show an improvement in terms of RMS values of prediction error. We also present an ablative analysis of the components of our proposed model and analyze the predictions made by the model in complex traffic scenarios.

Subfields: Artificial Intelligence, Automotive Engineering, Safety, Risk, Reliability and Quality

View Paper: https://www.semanticscholar.org/paper/305c4d91b0f70853a1cb0ed2a60a466b84e5c13d

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Automotive Engineering

In Safety, Risk, Reliability and Quality

Citations in the corpus

A Survey on Trajectory-Prediction Methods for Autonomous Driving (2022, methodology): “For trajectory prediction, combining CVAE and RNN variants into the form of encoder-decoder is an effective way for trajectory generation [75], [82], [84].”
Lidar for Autonomous Driving: The Principles, Challenges, and Trends for Automotive Lidar and Perception Systems (2020, background): “The authors in [41] proposed an encoder–decoder LSTM model to recognize maneuvers and predict trajectory.”
Attention Based Vehicle Trajectory Prediction (2020, methodology): “Some existing studies [4], [3], [5], [6] implicitly infer the dependencies between vehicles.”
Trajectron++: Dynamically-Feasible Trajectory Forecasting with Heterogeneous Data (2020, methodology): “On the nuScenes dataset, the following methods are also compared against: (6) Convolutional Social Pooling (CSP) [11]: An LSTM-based approach which explicitly considers a fixed number of movement classes and predicts which of those the modeled agent is likely to take.”
GRIP: Graph-based Interaction-aware Trajectory Prediction (2019, background): “Thus, instead of only predicting the position of one particular object as done in [9] and [11], the outputs Y of our proposed model are the predicted positions”
TrafficPredict: Trajectory Prediction for Heterogeneous Traffic-Agents (2018, methodology): “There are many classical approaches, including Bayesian networks (Lef`evre et al. 2011), Monte Carlo Simulation (Danielsson et al. 2007), Hidden Markov Models (HMM) (Firl et al. 2012), Kalman Filters (Kalman 1960), linear and non-linear Gaussian Process regression models (Rasmussen and Williams 2006”
Human motion trajectory prediction: a survey (2019, methodology): “…vehicle prediction methods consider interaction between trafﬁc participants, e.g. (Agamennoni et al. 2012; Kuhnt et al. 2016; Raipuria et al. 2018; Deo and Trivedi 2018; Kim et al. 2017; Broadhurst et al. 2005; K ¨ afer et al. 2010; Bahram et al. 2016; Srikanth et al. 2019; Park et al. 2018;…”
Multimodal Trajectory Predictions for Autonomous Driving using Deep Convolutional Networks (2018, methodology): “[38] the authors introduced a method for multimodal trajectory prediction for highway vehicles by learning a model that assigns probabilities to six maneuver classes.”
Deep Learning-Based Vehicle Behavior Prediction for Autonomous Driving Applications: A Review (2020, methodology): “M-LSTM [24] Track history of the TV and SVs Multimodal Trajectory RNN (Multiple RNNs) 0.”
GOHOME: Graph-Oriented Heatmap Output for future Motion Estimation (2021, background): “It translates into a spread of the possible future position and is often tackled by the use of Gaussian predictions in motion estimation [1, 2].”
Survey of Deep Reinforcement Learning for Motion Planning of Autonomous Vehicles (2020, background): “This approach can also be used representing probabilistic maneuvers of surrounding vehicles [63], or by generating spatiotemporal map from a predicted sequence of movements, motion planning in a dynamic environment can also be achieved [64].”
PRECOG: PREdiction Conditioned on Goals in Visual Multi-Agent Settings (2019, background): “In contrast to the commonly-used average displacement error (ADE) and final displacement error (FDE) metrics that computes the mean Euclidean error from a batch of samples to a single ground-truth sample [1, 6, 10, 14, 28], minMSD has the desirable property of not penalizing plausible samples that c”
Multi-Agent Tensor Fusion for Contextual Trajectory Prediction (2019, background): “Other approaches predict a set of possible trajectories, instead of a single deterministic trajectory, by conditioning on possible maneuver classes [9, 10].”
Naturalistic Driver Intention and Path Prediction Using Recurrent Neural Networks (2018, background): “Recently, there has been work in this area with a focus on vehicles on highways using processed visual data of surrounding vehicles [12], radar data [13] or simulated lidar returns of surround vehicles [14].”