F1/10 Team - KTH SML

Motivation for the Autonomous Racing Competition

Both in commercial and private use autonomous systems have become one of the most intriguing field of study. With continuing advancements of vehicles, buses, trucks and delivery drones, the presence of autonomous systems in our everyday life is increasing with an impressive rate.

With the ambivalent applications of autonomous algorithms, the skills to develop, design and implement autonomous systems pave the way for a very successful and interesting career.

Competition Rules

The competition rules have been published in the beginning of March (link) and the race day takes place during the Cyber Physical Systems week 2018 (CPSweek2018) in Porto, Portugal on the 10th - 13th April 2018.

In the course of the F1/10 competition, each team implements an autonomous system using a platform defined by the event organizer. The vehicle to be used is a 1:10 scale rally car that has been adapted with sensors to facilitate autonomous driving.

The range of sensors consists of a Hokuyo 10LX (2d-lidar), a stereo ZED camera and an Inertial Measurement Unit (IMU). For tackling the computational load of the algorithms, the teams can choose to utilise either the NVIDIA Tegra K1, X1 or X2.

Furthermore, does the event organizer permit the teams to adapt the vehicle by adjusting kinematic parameters within the frame of certain basic range definitions for its vehicle dimension and mass.

Educational Value

In the course of this competition members of this team make experiences with state-of-the-art project management methods. The organization is based on scrum has a flat hierarchy, daily internal feedback loops and 1h weekly external feedback loops with the supervising professor. (Jonas Mårtensson - Contact)
Additionally, several tools such as slack, trello, drive and github are incorporated for an optimized workflow.

The final implementation of the autonomous algorithm is done on the vehicle utilizing a NVIDIA Tegra K1. Herewith, the Robotic Operating System (ROS) constitutes the main framework for merging the different components with a common interface and easily defined transmissions.

In the process of the competition all team members obtain valuable skills to develop, design and implement algorithms in both python and c++.

Depending on the specific assignment to one of the corresponding groups for project management, perception and localisation, path planning or control and modelling, additional skills are developed alongside.

Group related skills:

• Perception and Localisation:
  → the focus is on the implementation of an algorithm for odometry, mapping and localisation based on monte carlo localization. (+obstacle avoidance)

• Path Planning:
  → the main challenge to be tackled is the real-time implementation given respective hardware restrictions, consideration of obstacles and the definition of the ideal race line

• Modelling and Control:
  →  the identification of vehicle parameters, the modelling of the dynamics and the development of the control via constrained model predictive control with reference tracking is the main task.

• Project Management:
  → focuses on the facilitation of an optimal work frame, the implementation of scrum feedback loops, documentation standards for individual progress, the ROS network framework, support for data acquisition / publishing for each team, transform, gazebo simulation and Github management.

Our Sponsors

We want to thank the Integrated Transport Research Lab (ITRL) for its support and provision of the Smart Mobility Lab (SML) as a testing site with its excellent qualisys (mocap) measurement system as ground truth for parameter identification. Furthermore, did we receive a preliminary funding for a part of the team to travel to the final competition location.

 Sponsorlist:

At the moment we are still looking for additional funding. Our goal is to provide every team member with transportation to the competition. Moreover, we would like to improve the platform with a NVIDIA Tegra X2 and some additional team related acquisitions depending on the available budget.