KTH in AD-EYE; Digitization of KTH main campus.
This video shows the various features of AD-EYE and testing in a digitized KTH.
Here, we replicated KTH buildings and roads using data from Open Street Maps and automated the process of creating multiple representations of the world used for our Automated Driving stack. This represents our capacity to run our tests in simulation of almost any real-world location , to test our code under various scenarios. A critical step before any real-world driving can take place.
AD-EYE on Road
This video shows the results of the students from the "AD-EYE on Road" group in the Capstone course MF2059 of the Mechatronics Master's track at KTH.
The video details the first steps for AD-EYE to transition from Simulation to Real-world in the quest to make KTH a testbed for ITS systems. In this video seen are points such as
The mapping of the KTH campus
Progress towards improving the digital twin of KTH in AD-EYE
The usage of the AD-EYE platform with the RCV-E
Safety Channel takeover
This video shows the Nominal Channel being interrupted by an injected fault in one of the Object Detection functions and:
1. The takeover by the Safety Channel and the execution of a Minimal Risk Condition i.e. a safe stop outside of the road,
2. Restart of the Nominal Channel and the resumption of the mission.
3D mapping of KTH Brinellvägen and Drottning Kristinas Väg
The video shows a 3d mapping (Lidar pointcloud) of a part of KTH main campus in Stockholm. This marks the start of our effort to create a digital twin of KTH, and simulation of real world locations.
Based on data collected by ITRL, and work by Florent Charton.
Safety Channel takeover : Research Concept Vehicle
This video demonstrates the execution of the takeover and the execution of a Minimal Risk Condition a safe stop outside the road, using the KTH Research Concept Vehicle.
Based on work primarily performed by KTH researchers Erik Ward and Lars Svensson.
A demo video made by the students of ESHK 2018: Group Nvidia using the AD-EYE platform demonstrating the first iteration of the Vector Mapping (Road network definition format used by Autoware) feature, and the ease of creating "Simulation Worlds" in AD-EYE
Open Scenario Implementation within AD-EYE
This video shows the implementation of an importing mechanism for scenarios based on the Open Scenario format, using parameter sweeps with the Test Automation feature (to control for different conditions such as Rain intensity), and evauation of the results.
Based on the work of Maurits Gernaat during his internship (2019)
AD-EYE improvements and Open Scenario based experiments
Final presentation of the work performed by Florent Charton as part of his 5 month internship at KTH Mechatronics. (2020).
The video shows the generality of the platform ADI, with the same code running fully automated in some of our many "Simulation Worlds" under various environmental conditions.
Based on the work of Radhika Tekade (VNIT) during her internship (2019)
CNN based classification
This video shows:
1) The generation of a comprehensive image dataset from PreScan world .
2) The training of the Convolutional Neural Network used with the SSD algorithm.
3) The improvements achieved in classification by the newly trained network, in comparision with a pretrained CNN.
This highlights the capacity to adapt CNNs within AD-EYE for the flexibility needed between different Automated Driving ODDs.
Based on the work of Paul Thomas (ENSMA) during his internship (2019)
Vector Map generation
This video shows the vast improvments made in the Vector Mapping feature of AD-EYE, and enables the automation of import road network definitions from real and simulated "Worlds".
Based on the work of Nicolas Helleboid (ENSMA) during his internship (2019)