ICRA 2020

We’re continuing our strong presence at the IEEE International Conference on Robotics and Automation (ICRA) in 2020, with five accepted papers! For the papers we presented in 2019, check out our ICRA 2019 page.

Our Papers with 10-minute Summary Videos

Context-aware cost shaping to reduce the impact of model error in safe, receding horizon control
C. D. McKinnon and A. P. Schoellig
in Proc. of the IEEE International Conference on Robotics and Automation (ICRA), 2020. Accepted.
[View BibTeX] [View Abstract] [Download PDF]

This paper presents a method to enable a robot using stochastic Model Predictive Control (MPC) to achieve high performance on a repetitive path-following task. In particular, we consider the case where the accuracy of the model for robot dynamics varies significantly over the path–motivated by the fact that the models used in MPC must be computationally efficient, which limits their expressive power. Our approach is based on correcting the cost predicted using a simple learned dynamics model over the MPC horizon. This discourages the controller from taking actions that lead to higher cost than would have been predicted using the dynamics model. In addition, stochastic MPC provides a quantitative measure of safety by limiting the probability of violating state and input constraints over the prediction horizon. Our approach is unique in that it combines both online model learning and cost learning over the prediction horizon and is geared towards operating a robot in changing conditions. We demonstrate our algorithm in simulation and experiment on a ground robot that uses a stereo camera for localization.

@INPROCEEDINGS{mckinnon-icra20,
title = {Context-aware Cost Shaping to Reduce the Impact of Model Error in Safe, Receding Horizon Control},
author = {Christopher D. McKinnon and Angela P. Schoellig},
booktitle = {{Proc. of the IEEE International Conference on Robotics and Automation (ICRA)}},
year = {2020},
note = {Accepted},
abstract = {This paper presents a method to enable a robot using stochastic Model Predictive Control (MPC) to achieve high performance on a repetitive path-following task. In particular, we consider the case where the accuracy of the model for robot dynamics varies significantly over the path–motivated by the fact that the models used in MPC must be computationally efficient, which limits their expressive power. Our approach is based on correcting the cost predicted using a simple learned dynamics model over the MPC horizon. This discourages the controller from taking actions that lead to higher cost than would have been predicted using the dynamics model. In addition, stochastic MPC provides a quantitative measure of safety by limiting the probability of violating state and input constraints over the prediction horizon. Our approach is unique in that it combines both online model learning and cost learning over the prediction horizon and is geared towards operating a robot in changing conditions. We demonstrate our algorithm in simulation and experiment on a ground robot that uses a stereo camera for localization.}
}

Slack Channel: #moc08_6


[DOI] Online trajectory generation with distributed model predictive control for multi-robot motion planning
C. E. Luis, M. Vukosavljev, and A. P. Schoellig
IEEE Robotics and Automation Letters, vol. 5, iss. 2, p. 604–611, 2020.
[View BibTeX] [View Abstract] [Download PDF] [View Video] [More Information]

We present a distributed model predictive control (DMPC) algorithm to generate trajectories in real-time for multiple robots. We adopted the on-demand collision avoidance method presented in previous work to efficiently compute non-colliding trajectories in transition tasks. An event-triggered replanning strategy is proposed to account for disturbances. Our simulation results show that the proposed collision avoidance method can reduce, on average, around 50\% of the travel time required to complete a multi-agent point-to-point transition when compared to the well-studied Buffered Voronoi Cells (BVC) approach. Additionally, it shows a higher success rate in transition tasks with a high density of agents, with more than 90\% success rate with 30 palm-sized quadrotor agents in a 18 m^3 arena. The approach was experimentally validated with a swarm of up to 20 drones flying in close proximity.

@article{luis-ral20,
title = {Online Trajectory Generation with Distributed Model Predictive Control for Multi-Robot Motion Planning},
author = {Carlos E. Luis and Marijan Vukosavljev and Angela P. Schoellig},
journal = {{IEEE Robotics and Automation Letters}},
year = {2020},
volume = {5},
number = {2},
pages = {604--611},
doi = {10.1109/LRA.2020.2964159},
urlvideo = {https://www.youtube.com/watch?v=N4rWiraIU2k},
urllink = {https://arxiv.org/pdf/1909.05150.pdf},
abstract = {We present a distributed model predictive control (DMPC) algorithm to generate trajectories in real-time for multiple robots. We adopted the on-demand collision avoidance method presented in previous work to efficiently compute non-colliding trajectories in transition tasks. An event-triggered replanning strategy is proposed to account for disturbances. Our simulation results show that the proposed collision avoidance method can reduce, on average, around 50\% of the travel time required to complete a multi-agent point-to-point transition when compared to the well-studied Buffered Voronoi Cells (BVC) approach. Additionally, it shows a higher success rate in transition tasks with a high density of agents, with more than 90\% success rate with 30 palm-sized quadrotor agents in a 18 m^3 arena. The approach was experimentally validated with a swarm of up to 20 drones flying in close proximity.}
}

Slack Channel: #mob11_1


[DOI] A data-driven motion prior for continuous-time trajectory estimation on SE(3)
J. N. Wong, D. J. Yoon, A. P. Schoellig, and T. D. Barfoot
IEEE Robotics and Automation Letters, vol. 5, iss. 2, p. 1429–1436, 2020.
[View BibTeX] [View Abstract] [Download PDF]

Simultaneous trajectory estimation and mapping (STEAM) is a method for continuous-time trajectory estimation in which the trajectory is represented as a Gaussian Process (GP). Previous formulations of STEAM used a GP prior that assumed either white-noise-on-acceleration (WNOA) or white-noise-on-jerk (WNOJ). However, previous work did not provide a principled way to choose the continuous-time motion prior or its parameters on a real robotic system. This paper derives a novel data-driven motion prior where ground truth trajectories of a moving robot are used to train a motion prior that better represents the robot’s motion. In this approach, we use a prior where latent accelerations are represented as a GP with a Matérn covariance function and draw a connection to the Singer acceleration model. We then formulate a variation of STEAM using this new prior. We train the WNOA, WNOJ, and our new latent-force prior and evaluate their performance in the context of both lidar localization and lidar odometry of a car driving along a 20km route, where we show improved state estimates compared to the two previous formulations.

@article{wong-ral20,
title = {A Data-Driven Motion Prior for Continuous-Time Trajectory Estimation on {SE(3)}},
author = {Jeremy N. Wong and David J. Yoon and Angela P. Schoellig and Timothy D. Barfoot},
journal = {{IEEE Robotics and Automation Letters}},
year = {2020},
volume = {5},
number = {2},
pages = {1429--1436},
doi = {10.1109/LRA.2020.2969153},
abstract = {Simultaneous trajectory estimation and mapping (STEAM) is a method for continuous-time trajectory estimation in which the trajectory is represented as a Gaussian Process (GP). Previous formulations of STEAM used a GP prior that assumed either white-noise-on-acceleration (WNOA) or white-noise-on-jerk (WNOJ). However, previous work did not provide a principled way to choose the continuous-time motion prior or its parameters on a real robotic system. This paper derives a novel data-driven motion prior where ground truth trajectories of a moving robot are used to train a motion prior that better represents the robot's motion. In this approach, we use a prior where latent accelerations are represented as a GP with a Mat\'{e}rn covariance function and draw a connection to the Singer acceleration model. We then formulate a variation of STEAM using this new prior.
We train the WNOA, WNOJ, and our new latent-force prior and evaluate their performance in the context of both lidar localization and lidar odometry of a car driving along a 20km route, where we show improved state estimates compared to the two previous formulations.}
}

Slack Channel: #tud07_02


Visual localization with Google Earth images for robust global pose estimation of UAVs
B. Patel, T. D. Barfoot, and A. P. Schoellig
in Proc. of the IEEE International Conference on Robotics and Automation (ICRA), 2020. Accepted.
[View BibTeX] [View Abstract] [Download PDF] [View Video]

We estimate the global pose of a multirotor UAV by visually localizing images captured during a flight with Google Earth images pre-rendered from known poses. We metrically localize real images with georeferenced rendered images using a dense mutual information technique to allow accurate global pose estimation in outdoor GPS-denied environments. We show the ability to consistently localize throughout a sunny summer day despite major lighting changes while demonstrating that a typical feature-based localizer struggles under the same conditions. Successful image registrations are used as measurements in a filtering framework to apply corrections to the pose estimated by a gimballed visual odometry pipeline. We achieve less than 1 metre and 1 degree RMSE on a 303 metre flight and less than 3 metres and 3 degrees RMSE on six 1132 metre flights as low as 36 metres above ground level conducted at different times of the day from sunrise to sunset.

@INPROCEEDINGS{patel-icra20,
title = {Visual Localization with {Google Earth} Images for Robust Global Pose Estimation of {UAV}s},
author = {Bhavit Patel and Timothy D. Barfoot and Angela P. Schoellig},
booktitle = {{Proc. of the IEEE International Conference on Robotics and Automation (ICRA)}},
year = {2020},
note = {Accepted},
urlvideo = {https://tiny.cc/GElocalization},
abstract = {We estimate the global pose of a multirotor UAV by visually localizing images captured during a flight with Google Earth images pre-rendered from known poses. We metrically localize real images with georeferenced rendered images using a dense mutual information technique to allow accurate global pose estimation in outdoor GPS-denied environments. We show the ability to consistently localize throughout a sunny summer day despite major lighting changes while demonstrating that a typical feature-based localizer struggles under the same conditions. Successful image registrations are used as measurements in a filtering framework to apply corrections to the pose estimated by a gimballed visual odometry pipeline. We achieve less than 1 metre and 1 degree RMSE on a 303 metre flight and less than 3 metres and 3 degrees RMSE on six 1132 metre flights as low as 36 metres above ground level conducted at different times of the day from sunrise to sunset.}
}

Slack Channel: #tuc07_6


Experience selection using dynamics similarity for efficient multi-source transfer learning between robots
M. J. Sorocky, S. Zhou, and A. P. Schoellig
in Proc. of the IEEE International Conference on Robotics and Automation (ICRA), 2020. Accepted.
[View BibTeX] [View Abstract] [Download PDF] [View Video]

In the robotics literature, different knowledge transfer approaches have been proposed to leverage the experience from a source task or robot—real or virtual—to accelerate the learning process on a new task or robot. A commonly made but infrequently examined assumption is that incorporating experience from a source task or robot will be beneficial. For practical applications, inappropriate knowledge transfer can result in negative transfer or unsafe behaviour. In this work, inspired by a system gap metric from robust control theory, the nu-gap, we present a data-efficient algorithm for estimating the similarity between pairs of robot systems. In a multi-source inter-robot transfer learning setup, we show that this similarity metric allows us to predict relative transfer performance and thus informatively select experiences from a source robot before knowledge transfer. We demonstrate our approach with quadrotor experiments, where we transfer an inverse dynamics model from a real or virtual source quadrotor to enhance the tracking performance of a target quadrotor on arbitrary hand-drawn trajectories. We show that selecting experiences based on the proposed similarity metric effectively facilitates the learning of the target quadrotor, improving performance by 62\% compared to a poorly selected experience.

@INPROCEEDINGS{sorocky-icra20,
author = {Michael J. Sorocky and Siqi Zhou and Angela P. Schoellig},
title = {Experience Selection Using Dynamics Similarity for Efficient Multi-Source Transfer Learning Between Robots},
booktitle = {{Proc. of the IEEE International Conference on Robotics and Automation (ICRA)}},
year = {2020},
note = {Accepted},
urlvideo = {https://youtu.be/8m3mOkljujM},
abstract = {In the robotics literature, different knowledge transfer approaches have been proposed to leverage the experience from a source task or robot—real or virtual—to accelerate the learning process on a new task or robot. A commonly made but infrequently examined assumption is that incorporating experience from a source task or robot will be beneficial. For practical applications, inappropriate knowledge transfer can result in negative transfer or unsafe behaviour. In this work, inspired by a system gap metric from robust control theory, the nu-gap, we present a data-efficient algorithm for estimating the similarity between pairs of robot systems. In a multi-source inter-robot transfer learning setup, we show that this similarity metric allows us to predict relative transfer performance and thus informatively select experiences from a source robot before knowledge transfer. We demonstrate our approach with quadrotor experiments, where we transfer an inverse dynamics model from a real or virtual source quadrotor to enhance the tracking performance of a target quadrotor on arbitrary hand-drawn trajectories. We show that selecting experiences based on the proposed similarity metric effectively facilitates the learning of the target quadrotor, improving performance by 62\% compared to a poorly selected experience.},
}

Slack Channel: #moc17_4


University of Toronto Institute for Aerospace Studies