6th Annual CROSS Research Symposium
October 12th, 2021
Workshop: Special Topics in Autonomous Systems
The workshop will feature an invited presentation by Addisu Taddese of Open Robotics on Ignition-Gazebo, an open source robotics simulator that provides high fidelity physics, rendering, and sensor models.
Additionally we feature several topics of interest for autonomous systems including the following:
Simulation of Autonomous Systems in Ignition-Gazebo
Addisu Taddese, Open Robotics
Abstract: Ignition-gazebo—a successor and rewrite of Gazebo (classic)—is an open source dynamic simulator with the goal of providing accurate physics simulation, realistic rendering, and high fidelity sensing while being highly modular and extensible. It is built using Ignition, a set of libraries for the design, development, and testing of robotics software aided by simulation. This talk will give an overview on how users can simulate autonomous systems in ignition-gazebo using the rich suite of sensors and the collection of ground, aerial, and marine vehicles that are readily available as part of ignition. It will also briefly cover how users can extend ignition-gazebo to create custom vehicles, sensors, and environments to suit their needs.
Addisu Taddese has been with Open Robotics since 2018. He's been actively involved in the development of several libraries that make up Ignition. In particular, he has made significant contributions to the integration of the DART physics engine into ignition-gazebo and currently helps maintain ignition-physics. He's also been a core developer in the recent improvements to SDFormat that allow users to leverage frame semantics and model composition when creating robot models. Addisu received his Ph.D. degree in Electrical Engineering from Vanderbilt University with a dissertation on magnetic pose estimation and robotic manipulation of magnetically actuated capsule endoscopes
IMU Calibration and Self-correction Algorithm for Open-source Autonomous Vehicle Controller
Rishikesh Vanarse, Carnegie Mellon
Abstract: An Inertial Measurement Unit (IMU) that measures the linear acceleration and angular velocity of a robot/vehicle is constrained due to physical limitations such as noise, drifts, misalignments and offsets. In order to minimize the drift in state estimates over time, accurate calibration is required.
The key contribution of this project was the development of a method for continuous real-time calibration and self-correction for an IMU, which quickly adapts to changing parameters. This method was tested with simulated as well as real data. The second contribution of this project was a comparison study of attitude estimation algorithms that fuse data from an accelerometer and magnetometer to estimate a robot’s orientation.
Rishikesh Vanarse completed his BS in Computer Science from BITS Pilani Goa, India, in 2021. His undergraduate thesis was on underwater SLAM under Dr Kostas Alexis and worked on multispectral drone imagery as an undergraduate researcher in The Autonomous Robots Lab at UNR/NTNU (2020-2021). Additionally, he taught and designed robotics courses at BITS Goa. he is currently an associate researcher at Carnegie Mellon in the Biorobotics Lab under Dr. Matt Travers and Dr. Howie Choset
Detecting Course Markers for Autonomous Vehicles
Rupal Sharma, IIT Baharas
Abstract: Briefly, the problem is detecting the course markers to help an autonomous vehicle find an optimal trajectory to complete the circuit.
In this project, I have used deep learning and computer vision(OpenCV) algorithms to detect the course markers.
To find an optimal path for the vehicle, we need to detect markers present in the circuit; thus, the coloured cone-like object will be spread throughout the course, both ML and CV algorithms will be deployed on Raspberry-Pi4.
The deep learning algorithms will find detected markers’ locations and use that; various information of the markers (distance, angle, etc.) relative to the vehicle will be calculated.
Rupal Sharma is in his senior year at The Indian Institute of Technology, Banaras Hindu University.
Partitioned Gaussian Process Regression for Online Trajectory Planning for Autonomous Vehicles
Pavlo Vlastos, UCSC
Abstract: Gaussian process regression and ordinary kriging are effective methods for spatial estimation, but are generally not used in online trajectory-planning applications for autonomous vehicles. A common use for kriging is spatial estimation for exploration. Kriging is limited by the necessary covariance matrix inversion and its computational complexity. Using the Sherman-Morison matrix inversion lemma, the complexity can be reduced. This work focuses on further improving the computational time required for spatial estimation with partitioned ordinary kriging (POK) for online trajectory-planning using the OSAVC.
Pavlo Vlastos is a PhD candidate at the University of California at Santa Cruz. He was born in Greece, grew up in Casper Wyoming, and moved to Santa Cruz to study and research autonomous systems. He is interested in designing and researching autonomous systems to reduce the cost of exploration and environmental study from oceans to space. His hobbies include biking and playing Kerbel Space Program.
Distributed Control Architecture for Resource-Constrained Autonomous Systems
Aaron Hunter, UCSC, CROSS Fellow
Abstract: We present a distributed architecture intended for deployment on small autonomous vehicles. This architecture consists of a real time controller, a guidance and navigation computer, and optionally an edge TPU for image classification. This architecture is deployed in several vehicles. We discuss some algorithms developed for this architecture for guidance, object detection, and attitude estimation.
Aaron Hunter is a PhD candidate at the University of California at Santa Cruz and a CROSS Fellow. His research concerns real time control of autonomous systems. In a previous incarnation, Aaron developed thermal processing equipment for the semiconductor industry. His interests are in outdoor activities of any kind, but preferably bicycle-related.