r2p2: A ReparameteRized Pushforward Policy for Diverse, Precise Generative Path Forecasting

By Nicholas Rhinehart, Kris M. Kitani, Paul Vernaza. 2018

257 citations recorded by the source corpus.

Subfields: Civil and Structural Engineering, Signal Processing

How researchers see this paper

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

In Civil and Structural Engineering

In Signal Processing

Citations in the corpus

Deep Imitative Models for Flexible Inference, Planning, and Control (2018, methodology): “We modiﬁed the “RNN” method described by Rhinehart et al. (2018) and used LIDAR features χ = R 200 × 200 × 2 , with χ ij representing a 2-bin histogram of points below and above the ground in 0 .”
Imitation Learning as f-Divergence Minimization (2019, methodology): “Our framework encompasses methods that look at specific measures of divergence such as minimizing relative entropy [41] or symmetric crossentropy [42].”
Large Scale Interactive Motion Forecasting for Autonomous Driving : The Waymo Open Motion Dataset (2021, methodology): “For methods that provide explicit or implicit future probability distributions, the likelihood of the ground truth future trajectory can be used as a metric [8, 30, 27, 28].”
Argoverse: 3D Tracking and Forecasting With Rich Maps (2019, methodology): “al inter-vehicle interactions in the data. Another challenging problem in the trajectory forecasting domain is to predict diverse trajectories which can address multimodal nature of the problem. R2P2 [50] address the diversity-precision trade-off of generative forecasting models and formulate a symm”
MultiPath: Multiple Probabilistic Anchor Trajectory Hypotheses for Behavior Prediction (2019, methodology): “3 reproduces results reported by [18] for the DESIRE [3], SocialGAN [19], R2P2-MA [4], and the PRECOG-ESP [18] methods and compares the performance of MultiPath against them.”
DLow: Diversifying Latent Flows for Diverse Human Motion Prediction (2020, background): “problem in GANs [11, 12, 65, 4, 25, 16, 47] and the posterior collapse problem in VAEs [76, 66, 36, 8, 48, 29]. Normalizing ows [58] have also been used to promote diversity in trajectory forecasting [59, 24]. This line of work aims to improve the diversity of the data distribution learned by deep g”
HiVT: Hierarchical Vector Transformer for Multi-Agent Motion Prediction (2022, background): “The past trajectories of agents are either rasterized as additional image channels [9, 12] or processed by temporal models like RNNs [2, 41, 42].”
A Survey on Trajectory-Prediction Methods for Autonomous Driving (2022, methodology): “Some methods that use raw sensor data as input also use CVAE for multi-modal trajectory prediction [108], [156], [157].”
DenseTNT: End-to-end Trajectory Prediction from Dense Goal Sets (2021, background): “Some approaches aimed to address mode collapse [25, 36, 23, 10, 3].”
Of Moments and Matching: A Game-Theoretic Framework for Closing the Imitation Gap (2021, background): “2016), Jensen-Shannon (Rhinehart et al. 2018), Reverse KL (Kostrikov et al.”
Multi-Agent Tensor Fusion for Contextual Trajectory Prediction (2019, background): “Recent data-driven approaches based on deep networks [1, 4, 9, 10, 13, 19, 20, 24, 28, 29, 31] outperform traditional approaches.”
CoverNet: Multimodal Behavior Prediction Using Trajectory Sets (2019, background): “R2P2 [30] improves sample coverage for such policies via a symmetric KL loss.”
V2VNet: Vehicle-to-Vehicle Communication for Joint Perception and Prediction (2020, background): “Motion forecasting algorithm performance has been further improved by building good multimodal distributions [4,6,12,19] that capture diverse actor behaviour and by modelling actor interactions [3,25,36,37].”
PRECOG: PREdiction Conditioned on Goals in Visual Multi-Agent Settings (2019, methodology): “Our model is generative, and learns to map latent variables Z via an invertible function f to generate multi-agent state trajectories conditioned on φ . f ’s invertibility induces q ( S | φ ) , a pushforward distribution [23], also known as an invertible generative model [7, 12, 13, 16, 29].”
Motion Transformer with Global Intention Localization and Local Movement Refinement (2022, background): “Early works [1, 20, 39, 44, 40] propose to generate a set of trajectory samples to approximate the output distribution.”
MultiPath++: Efficient Information Fusion and Trajectory Aggregation for Behavior Prediction (2021, background): “In applications related to future prediction, capturing the highly uncertain and multimodal set of outcomes is a key challenge and the focus of much work [31, 37, 40, 45, 50].”
Human motion trajectory prediction: a survey (2019, methodology): “A semantic map (Karasev et al. 2016; Kitani et al. 2012; Rehder et al. 2018; Coscia et al. 2018; Shen et al. 2018; Vasishta et al. 2017, 2018; Rhinehart et al. 2018; Rhinehart et al. 2018; Tadokoro et al. 1993; Ballan et al. 2016; Zhao et al. 2019) or extracted features from an image (Xue et al.…”
AgentFormer: Agent-Aware Transformers for Socio-Temporal Multi-Agent Forecasting (2021, background): “As the future trajectory of an agent is uncertain and often multi-modal, recent trajectory prediction methods start to model the trajectory distribution with deep generative models [21, 13, 40] such as conditional variational autoencoders (CVAEs) [27, 57, 19, 46, 51, 45], generative adversarial netw”
MP3: A Unified Model to Map, Perceive, Predict and Plan (2021, methodology): “As there are multiple possible futures, these methods either generate a fixed set of trajectories [6, 8–10, 26, 28, 30, 36, 54], draw samples to characterize the distribution [7, 41, 47] or predict temporal occupancy maps [23, 27, 43].”