Multimodal Probabilistic Model-Based Planning for Human-Robot Interaction

By Edward Schmerling, Karen Leung, Wolf Vollprecht, Marco Pavone. 2017

187 citations recorded by the source corpus.

This paper presents a method for constructing human-robot interaction policies in settings where multimodality, i.e., the possibility of multiple highly distinct futures, plays a critical role in decision making. We are motivated in this work by the example of traffic weaving, e.g., at highway on-ramps/off-ramps, where entering and exiting cars must swap lanes in a short distance-a challenging negotiation even for experienced drivers due to the inherent multimodal uncertainty of who will pass whom. Our approach is to learn multimodal probability distributions over future human actions from a dataset of human-human exemplars and perform real-time robot policy construction in the resulting environment model through massively parallel sampling of human responses to candidate robot action sequences. Direct learning of these distributions is made possible by recent advances in the theory of conditional variational autoencoders (CVAEs), whereby we learn action distributions simultaneously conditioned on the present interaction history, as well as candidate future robot actions in order to take into account response dynamics. We demonstrate the efficacy of this approach with a human-in-the-loop simulation of a traffic weaving scenario.

How researchers see this paper

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

Citations in the corpus

Deep Imitative Models for Flexible Inference, Planning, and Control (2018, background): “Typically this involves recurrent neural networks combined with Gaussian density layers or generative models based on some context inputs such as LIDAR, images, or known positions of external agents (Lee et al., 2017; Schmerling et al., 2018; Zyner et al., 2018; Gupta et al., 2018; Ma et al., 2017).”
Learning Sampling Distributions for Robot Motion Planning (2017, methodology): “The data was collected from human demonstration on a two person driving simulator [32].”
Human motion trajectory prediction: a survey (2019, methodology): “…approaches based on their main modeling method. iii) Methods that use behavior cloning (imitation of human behaviors with supervised learning techniques), i.e. learn/recover the motion model directly from data, are classiﬁed as pattern-based approaches (Schmerling et al. 2018; Zheng et al. 2016).”
Learning with Humans in the Loop (2020, methodology): “, a conditional neural network, as in [52].”
Naturalistic Driver Intention and Path Prediction Using Recurrent Neural Networks (2018, methodology): “[16] that use a Conditional Variational Autoencoder will produce a prediction without uncertainty, but the model may be sampled many times to produce multiple predictions, which may be combined to estimate uncertainty.”
PRECOG: PREdiction Conditioned on Goals in Visual Multi-Agent Settings (2019, methodology): “In terms of multi-agent forecasting, our work is re-lated to [31] which uses a conditional VAE [17] encoding of the joint states of multiple agents together with recurrent cells to predict future human actions.”
The Trajectron: Probabilistic Multi-Agent Trajectory Modeling With Dynamic Spatiotemporal Graphs (2018, background): “z’s purpose is to model latent structure in the interaction which both improves learning performance and enables interpretation of results [24, 43, 46].”