• Corpus ID: 195886415

Kernel Trajectory Maps for Multi-Modal Probabilistic Motion Prediction

@article{Zhi2019KernelTM,
  title={Kernel Trajectory Maps for Multi-Modal Probabilistic Motion Prediction},
  author={Weiming Zhi and Lionel Ott and Fabio Tozeto Ramos},
  journal={ArXiv},
  year={2019},
  volume={abs/1907.05127}
}
Understanding the dynamics of an environment, such as the movement of humans and vehicles, is crucial for agents to achieve long-term autonomy in urban environments. This requires the development of methods to capture the multi-modal and probabilistic nature of motion patterns. We present Kernel Trajectory Maps (KTM) to capture the trajectories of movement in an environment. KTMs leverage the expressiveness of kernels from non-parametric modelling by projecting input trajectories onto a set of… 

Figures and Tables from this paper

Probabilistic Trajectory Prediction with Structural Constraints

This work addresses the problem of predicting the motion trajectories of dynamic objects in the environment with a novel framework, which combines probabilistic learning and constrained trajectory optimisation, and focuses on collision constraints, such that extrapolated future trajectory distributions conform to the environment structure.

Trajectory Generation in New Environments from Past Experiences

The Occupancy-Conditional Trajectory Network (OTNet) framework is introduced, capable of transferring the previously observed motion patterns in known environments to new environments, and its ability to generate complex multi-modal trajectory patterns in both simulated and real-world environments is demonstrated.

Social-VRNN: One-Shot Multi-modal Trajectory Prediction for Interacting Pedestrians

This paper presents a variational learning approach for interaction-aware and multi-modal trajectory prediction based on deep generative neural networks that can achieve faster convergence and requires significantly fewer samples comparing to state-of-the-art methods.

Anticipatory Navigation in Crowds by Probabilistic Prediction of Pedestrian Future Movements

This paper presents Stochastic Process Anticipatory Navigation (SPAN), a framework that enables nonholonomic robots to navigate in environments with crowds, while anticipating and accounting for the motion patterns of pedestrians.

Probabilistic Crowd GAN: Multimodal Pedestrian Trajectory Prediction Using a Graph Vehicle-Pedestrian Attention Network

Improvements on the existing state of the art methods for trajectory prediction are demonstrated and the true multimodal and uncertain nature of crowd interactions can be directly modelled are illustrated.

Deep Learning for Vision-based Prediction: A Survey

An overview of the vision-based prediction field in the past five years with a particular focus on deep learning approaches is provided, categorizing these algorithms into video prediction, action prediction, trajectory prediction, body motion prediction, and other prediction applications.

References

SHOWING 1-10 OF 37 REFERENCES

Spatiotemporal Learning of Directional Uncertainty in Urban Environments With Kernel Recurrent Mixture Density Networks

A method to build a continuous map to capture spatiotemporal movements of a typical object in the environment over time, which provides a multi-modal probability density function over the possible directions an object can move in.

Modelling and Predicting Rhythmic Flow Patterns in Dynamic Environments

A time-dependent probabilistic map able to model and predict flow patterns of people in indoor environments and potential applications include human-aware motion planning, improving the efficiency and safety of robot navigation.

Conditional transition maps: Learning motion patterns in dynamic environments

A spatial representation, which encapsulates the statistical dynamic behavior observed in the environment, and a proof-of-concept method for estimating future paths of dynamic objects, called Conditional Probability Propagation Tree (CPPTree).

Multi-modal estimation with kernel embeddings for learning motion models

This work presents a novel estimation algorithm for filtering and regression with a number of advantages over existing methods, a generalisation of the Kernel Bayes' Rule that produces multi-modal posterior estimates represented as Gaussian mixtures.

Continuous-time Gaussian process motion planning via probabilistic inference

This work introduces a novel formulation of motion planning, for continuous-time trajectories, as probabilistic inference, and develops a very efficient algorithm that combines GP representations of trajectories with fast, structure-exploiting inference via numerical optimization.

Modeling motion patterns of dynamic objects by IOHMM

A novel approach to model motion patterns of dynamic objects in the environment with the occupancy grid map representation by introducing observations of neighbor cells in the previous time step as input of IOHMM, which enables the method to model the spatial correlation of dynamics across cells.

Directional Grid Maps: Modeling Multimodal Angular Uncertainty in Dynamic Environments

Directional statistics into robotic mapping to model circular data is introduced and directional grid maps are proposed to represent the location-wide long-term angular motion of the environment to improve robustness, efficiency, and safety.

Stochastic functional gradient for motion planning in continuous occupancy maps

This work introduces a novel planning method based on trajectory optimization to plan safe and efficient paths in continuous occupancy maps by devising a stochastic gradient update rule to optimize a path represented as a Gaussian process.

Functional Gradient Motion Planning in Reproducing Kernel Hilbert Spaces

This work generalizes functional gradient trajectory optimization by formulating it as minimization of a cost functional in an RKHS, which lets it represent trajectories as linear combinations of kernel functions, without any need for waypoints.

Batch Continuous-Time Trajectory Estimation as Exactly Sparse Gaussian Process Regression

This paper revisits batch state estimation through the lens of Gaussian process (GP) regression, and shows that this class of prior results in an inverse kernel matrix that is exactly sparse (block-tridiagonal) and that this can be exploited to carry out GP regression (and interpolation) very efficiently.