From Friday, April 19th (11:00 PM CDT) through Saturday, April 20th (2:00 PM CDT), 2024, ni.com will undergo system upgrades that may result in temporary service interruption.
We appreciate your patience as we improve our online experience.
Student Projects
Contact Information
University: Swiss Federal Institute of Technology (ETH) Zurich
Competition Year: 2017/2018
Team Members: Nikhilesh Alatur, Nicholas Bünger, Robin Deuber, Frederick Gonon,
Niklas Funk, Nico Messikommer, Julian Nubert, Moritz Patriarca,
Simon Schaefer, Dominic Scotoni
Faculty Advisers: Prof. Dr. Roland Siegwart, Renaud Dubé, Rhagav Khanna, Dr. Philipp Krüsi, Mark Pfeiffer
Email Address: rdeuber@ethz.ch
Website: Project ARC
Language English
Project Information
Title
Project ARC – Development of an Autonomous Electric Race Car
Description
We were the first team to develop an autonomous car capable of cresting a Swiss mountain pass using solely a stereo-camera for localization. With our project, we wanted to increase the enthusiasm of the public for autonomous driving and show its feasibility!
Products
NI CompactRIO-9066 with modules NI 9381, NI 9269, NI 9881, NI 9403 and LabVIEW 2016
The Challenge
Autonomous driving and electric vehicles are nowadays highly active research and development areas. In this spirit, the Project ARC dealt with the conversion of an electric car, an eRod from the manufacturer Kyburz, into an autonomous vehicle that is capable of operating in challenging unstructured environments.
Our vision was to conquer a Swiss mountain road autonomously using a vision-based teach and repeat method. During one year, our project was supervised and supported by Prof. Dr. Roland Siegwart from the Autonomous Systems Lab Zurich. Our team consisted of eight mechanical and two electrical engineering undergraduate students who were all in their third year at Swiss Federal Institute of Technology (ETH) Zurich.
The team of Project ARC and its autonomous car.
The Solution
Project Overview - the finally resulting performance of our vehicle:
Platform Overview
The eRod is an all-electric vehicle produced by Kyburz Switzerland and was provided to us as basis for our development of a self-driving vehicle. In order to control the vehicle via digital commands, new actuators had to be implemented. Namely, a new steering unit provided by Thyssenkrupp Presta, which is controlled via CAN, was incorporated as well as a new braking system. The CAD model of the modified eRod is shown in the following figure.
The CAD model of the ARC eRod with its additional sensors and the computer in red color.
Sensors
Our localization algorithms rely on the input from the Visual Inertial sensor (VI sensor) which is mounted on the bonnet. This sensor module contains a stereo camera together with an inertial measurement unit (IMU). A LiDAR sensor, which is mounted on top of the roll-over bar, was used for the obstacle detection. The on-board sensor system also includes two rotary wheel encoders which are read out using the FPGA on a National Instruments CompactRIO. All the sensor information is processed in the ROS framework running on our Linux-based computer. The whole system architecture is shown below.
Software system architecture of Project ARC.
System Architecture
Our autonomous approach is based on the teach and repeat method and is performed by a vision-based localization combined with control and path-planning algorithms. The computer is the core, taking care of state estimation, obstacle detection, and the computation of the high-level control commands, which are sent through our Interface to the CompactRIO, where the low-level controllers for velocity (acceleration and braking) and steering angle are implemented.
Since the CompactRIO controls the car’s velocity and is responsible for the low-level communication, it is considered the main VCU (Vehicle Control Unit). The CompactRIO takes control of the steering, permanently checks all UDP connections and inherits all the safety measures on the VCU level.
Safety
When developing an autonomous vehicle, our most important goal was to guarantee a maximum amount of safety at any point in time. Therefore we set up different safety layers and decided to test always with a person in the driver seat, who is able to take over the control of the eRod at any time.
The CompactRIO is the most essential hardware component in order to guarantee a safe and robust system. Due to our system architecture, all safety measures are performed by the CompactRIO as it controls the cars’ velocity and steering angle. In case of an emergency stop sent by the computer, an interface or computer breakdown, the steering intervention of a human or reaching the end of an autonomous drive, the velocity controller is disabled. The first four previously mentioned events also trigger the emergency braking procedure which causes a full stop of the vehicle.
The following video shows a situation in which suddenly obstacles occur in the planned path of the vehicle:
But not only in case of an emergency has safety to be ensured. The system, in the regular mode of operation, has to be robust and safe as well. There are many safety checks at the beginning of an autonomous drive, before the CompactRIO confirms the computer that it is ready to start driving. If one of the safety checks fails, no confirmation will be sent to the computer and the velocity controller will not start under these circumstances.
Benefits of using LabVIEW and NI tools
The main benefits of the CompactRIO for us was the high reliability of its real-time operating system and its clear structure. Using LabVIEW prevented us from developing other complicated embedded control systems. The intuitive LabVIEW user interface and the easy programming language allowed us to rapidly design and test the controlling programs. The flexibility of the modules used in combination with the CompactRIO had the benefit, that our configuration perfectly matched our requirements. Additionally, National Instruments not only sponsored us with high quality hardware, but also with very helpful support.
However, the main advantage was the very high reliability of the NI hard- and software. Due to the safety algorithms implemented on the CompactRIO, we never had any accidents, although our high-level control algorithms often failed during the testing phase.
Conclusion
With the help of National Instruments, ten undergraduate students were able to build a safe autonomous car within one year. The development of our low level control system would hardly be possible to such a degree of safety and robustness in this short time without the support of National Instruments. We are glad and proud to have this partnership with National Instruments.
