Predicting wide receiver trajectories in American football

Lee, Namhoon

doi:10.1109/wacv.2016.7477732

Cited by 54 publications

(40 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, if feature maps related to ARS are included, this problem could be resolved. Further, because animal trajectories might be affected by spatiotemporally changing landscape features that are both environmental (e.g., weather, wind, time of day, lunar rhythm; Baigas et al 2017) and social factors (e.g., the presence of congeners; Yoda et al 2011), IRL has the potential for increasing interpolation precision by incorporating these dynamically changing factors into the feature space (Lee and Kitani 2016). Ecologists now have access to various environmental data obtained from satellites (Running et al 2004, Neumann et al 2015 and drones (Turner 2014, Zhang et al 2016, Strandburg-Peshkin et al 2017, which could be incorporated as feature maps.…”

Section: Discussionmentioning

confidence: 99%

“…Ecologists now have access to various environmental data obtained from satellites (Running et al 2004, Neumann et al 2015 and drones (Turner 2014, Zhang et al 2016, Strandburg-Peshkin et al 2017, which could be incorporated as feature maps. Further, because animal trajectories might be affected by spatiotemporally changing landscape features that are both environmental (e.g., weather, wind, time of day, lunar rhythm; Baigas et al 2017) and social factors (e.g., the presence of congeners; Yoda et al 2011), IRL has the potential for increasing interpolation precision by incorporating these dynamically changing factors into the feature space (Lee and Kitani 2016). Because we extended the state space to three-dimensional grid cells, our interpolation method has much potential for considering dynamically changing factors.…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

Can AI predict animal movements? Filling gaps in animal trajectories using inverse reinforcement learning

et al. 2018

View full text Add to dashboard Cite

Focal animal sampling and continuous recording of behavior in situ are essential in the study of ecology. However, observation gaps and missing records are unavoidable because the focal individual can move out of sight and recording devices do not always work properly. Using an inverse reinforcement learning (IRL) framework, we have developed a novel gap-filling method to predict the most likely route that an animal would have traveled; within this framework, an algorithm learns a reward function from animal trajectories to find the environmental features preferred by the animal. We applied this approach to GPS trajectories obtained from streaked shearwaters (Calonectris leucomelas) and provide evidence of the advantages of the IRL approach over previously used interpolation methods. These advantages are as follows: (1) No assumptions about the parametric distribution governing movements are needed, (2) no assumptions regarding landscape preferences and restrictions are needed, and (3) large spatiotemporal gaps can be filled. This work demonstrates how IRL can enhance the ability to fill gaps in animal trajectories and construct reward-space maps in heterogeneous environments. The proposed methodology can assist movement research, which seeks to understand phenomena that are ecologically and evolutionarily significant, such as habitat selection and migration.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

Can AI predict animal movements? Filling gaps in animal trajectories using inverse reinforcement learning

et al. 2018

View full text Add to dashboard Cite

show abstract

“…The problem of trajectory prediction can be formulated as a reward-maximization problem in an IRL framework, and has been done so for American football player movements (rife with challenging interaction scenarios) [3]. It was shown that a player's path can be modeled as a Markov Decision Process (MDP) where the path reward is learned via maximum entropy IRL on a mixture of static hand-crafted and dynamic features.…”

Section: A Existing Methodsmentioning

confidence: 99%

Generative Modeling of Multimodal Multi-Human Behavior

Ivanovic

Szczerbicki

Leung

et al. 2018

2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)

View full text Add to dashboard Cite

This work presents a methodology for modeling and predicting human behavior in settings with N humans interacting in highly multimodal scenarios (i.e. where there are many possible highly-distinct futures). A motivating example includes robots interacting with humans in crowded environments, such as self-driving cars operating alongside humandriven vehicles or human-robot collaborative bin packing in a warehouse. Our approach to model human behavior in such uncertain environments is to model humans in the scene as nodes in a graphical model, with edges encoding relationships between them. For each human, we learn a multimodal probability distribution over future actions from a dataset of multi-human interactions. Learning such distributions is made possible by recent advances in the theory of conditional variational autoencoders and deep learning approximations of probabilistic graphical models. Specifically, we learn action distributions conditioned on interaction history, neighboring human behavior, and candidate future agent behavior in order to take into account response dynamics. We demonstrate the performance of such a modeling approach in modeling basketball player trajectories, a highly multimodal, multi-human scenario which serves as a proxy for many robotic applications.

show abstract

“…an attractive force towards their goal position and a repulsive force for other people, enabling collision avoidance). Since then, many other kinds of approaches have formulated trajectory forecasting as a sequence-modeling regression problem, and powerful approaches such as Inverse Reinforcement Learning (IRL) [32], Gaussian Process Regression (GPR) [10,39,50], and Recurrent Neural Networks (RNNs) [1,33,49] have been applied with strong performance. However, IRL mostly relies on a unimodal assumption of interaction outcome [28,34]; GPR falls prey to long inference times, rendering it infeasible for robotic usecases; and standard RNN methods cannot handle multimodal data.…”

Section: Related Workmentioning

confidence: 99%

The Trajectron: Probabilistic Multi-Agent Trajectory Modeling With Dynamic Spatiotemporal Graphs

Ivanovic

Pavone

2019

2019 IEEE/CVF International Conference on Computer Vision (ICCV)

404

307

View full text Add to dashboard Cite

Developing safe human-robot interaction systems is a necessary step towards the widespread integration of autonomous agents in society. A key component of such systems is the ability to reason about the many potential futures (e.g. trajectories) of other agents in the scene. Towards this end, we present the Trajectron, a graph-structured model that predicts many potential future trajectories of multiple agents simultaneously in both highly dynamic and multimodal scenarios (i.e. where the number of agents in the scene is time-varying and there are many possible highlydistinct futures for each agent). It combines tools from recurrent sequence modeling and variational deep generative modeling to produce a distribution of future trajectories for each agent in a scene. We demonstrate the performance of our model on several datasets, obtaining state-of-the-art results on standard trajectory prediction metrics as well as introducing a new metric for comparing models that output distributions.

show abstract

Predicting wide receiver trajectories in American football

Cited by 54 publications

References 23 publications

Can AI predict animal movements? Filling gaps in animal trajectories using inverse reinforcement learning

Can AI predict animal movements? Filling gaps in animal trajectories using inverse reinforcement learning

Generative Modeling of Multimodal Multi-Human Behavior

The Trajectron: Probabilistic Multi-Agent Trajectory Modeling With Dynamic Spatiotemporal Graphs

Contact Info

Product

Resources

About