gotthard driverless Overall car concept Formula Student is not - - PowerPoint PPT Presentation

Overall car concept

gotthard driverless

“Formula Student is not about designing good cars, it’s about designing good teams”

Claude Rouelle

12.05.2019 Akademischer Motorsportverein Zürich | Angelo Corradi 2

Agenda

• Team
• Methods
• Concept
• Implementation
• Discussion

12.05.2019 Akademischer Motorsportverein Zürich | Angelo Corradi 3

Courtesy of FSG

Team

12.05.2019 Akademischer Motorsportverein Zürich | Angelo Corradi 4

Management Hardware SW architecture Estimation Controls Perception

Team: Training

• Spread out skills in the team
• Focused workshops
• ROS
• Git
• C++
• But also:
• Rules (a lot!)
• Vehicle dynamics
• Basics of Formula Student
• CAD and FEM
• Safety procedures

12.05.2019 Akademischer Motorsportverein Zürich | Angelo Corradi 5

Courtesy of FSG and FS Austria

Methods: Team discipline

• Knowledge transfer tools
• Meeting records
• Alumni reviews
• Design mocks
• Tech inspections simulations
• Race camps

12.05.2019 Akademischer Motorsportverein Zürich | Angelo Corradi 6

Courtesy of ZF Friedrichshafen

Methods: Testing

• Efficiency: time, people, money
• Extensive use of simulation and

recorded data

• Test “on the real thing”
• Integration tests at home
• Easy visualization

12.05.2019 Akademischer Motorsportverein Zürich | Angelo Corradi 7

Concept: Team goals

• 1. Maximise points in a FSD event
• 2. Finish all disciplines
• 3. Safety is a priority
• 4. Solid foundation for AMZ driverless

12.05.2019 Akademischer Motorsportverein Zürich | Angelo Corradi 8

Courtesy of FSG

Concept

• Hardware reliability and integration
• Sensor redundancy
• Algorithm robustness
• Minimize lap time

12.05.2019 Akademischer Motorsportverein Zürich | Angelo Corradi 9

Courtesy of FSG

gotthard driverless

Akademischer Motorsportverein Zürich | Angelo Corradi

Courtesy of FSG

Base vehicle

12.05.2019 Akademischer Motorsportverein Zürich | Angelo Corradi 11

• CFRP one-piece monocoque
• Heave springs
• Air springs with MRF
• Chassis diffuser
• Split accumulator packaging
• Self-designed motors

Accumulator

• New accumulator needed
• Decided to rebuild only the

rear section

• 43% of the energy
• 17 Kg lighter

12.05.2019 Akademischer Motorsportverein Zürich | Angelo Corradi 12

Computing platform

12.05.2019 Akademischer Motorsportverein Zürich | Angelo Corradi 13

Slave Jetson TX2 Master MPL PIP 39 GPU Nvidia GTX 1050 Ti

Steering Actuator

• Direct load-path
• Steering rate covers 99% of driver speeds
• Design to withstand parksteer (peak) and

skidpad (continuous load)

12.05.2019 Akademischer Motorsportverein Zürich | Angeo Corradi 14

Emergency Braking System

• Normally braking configuration
• Hydro-pneumatic circuit
• Redundant and fail safe
• Compatible with a driver

No need for complex HW installation when switching to manual driving

12.05.2019 Akademischer Motorsportverein Zürich | Angelo Corradi 15

Software Concept

• Perception
• Vision
• Lidar
• Estimation
• Velocity Estimation
• SLAM
• Controls
• Discovery mode
• Race mode

12.05.2019 Akademischer Motorsportverein Zürich | Angelo Corradi 16

Perception: Sensor setup

• Cameras
• 3 Basler ace cameras:
• 2MP resolution
• GigE cameras
• CMOS sensor
• Global shutter
• Lidar
• Velodyne VLP-16 Hi-Res
• 16 Channels
• High range
• Low power consumption
• Up to 600,000 points per sec

12.05.2019 Akademischer Motorsportverein Zürich | Angelo Corradi 17

Perception: Mono setup

12.05.2019 Akademischer Motorsportverein Zürich | Angelo Corradi 18

Cone detection

Grid cell: 1m x 1m

Keypoint extraction 2D to 3D Position

Perception: Mono cone detection

12.05.2019 Akademischer Motorsportverein Zürich | Angelo Corrad 19

Perception: Stereo setup

12.05.2019 Akademischer Motorsportverein Zürich | Angelo Corradi 20

Cone detection Bounding box propagation Feature matching Triangulation

Grid cell: 1m x 1m

Keypoint extraction 2D to 3D

Perception: Stereo cone estimation

12.05.2019 Akademischer Motorsportverein Zürich | Angelo Corradi 21

Perception: Lidar

12.05.2019 Akademischer Motorsportverein Zürich | Angelo Corradi 22

Perception: Lidar visualization

12.05.2019 Akademischer Motorsportverein Zürich | Angelo Corradi 24

Velocity estimation: Sensor setup

12.05.2019 Akademischer Motorsportverein Zürich | Angelo Corradi 25

IMU + GNSS Wheel angular speed Motor Torque Optical speed sensor (Correvit) Antennas (dual setup) Steering angle

Estimation: Velocity estimation

12.05.2019 Akademischer Motorsportverein Zürich | Angelo Corradi 26

Estimation: SLAM

12.05.2019 Akademischer Motorsportverein Zürich | Angelo Corradi 27

Input Algorithm Output

Estimation: SLAM

12.05.2019 Akademischer Motorsportverein Zürich | Angelo Corradi 28

Controls: boundary estimation

12.05.2019 Akademischer Motorsportverein Zürich | Angelo Corradi 29

Controls: boundary estimation and MPCC

12.05.2019 Akademischer Motorsportverein Zürich | Angelo Corradi 30

12.05.2019 Akademischer Motorsportverein Zürich | Angelo Corradi 31

“Engineers like to solve problems. If there are no problems handily available, they create their own.”

Scott Adams

12.05.2019 Akademischer Motorsportverein Zürich | Angelo Corradi 32

FSD is Formula Student

• It’s not a software competition, it’s an engineering competition
• It retains all challenges of Formula Student
• Trade off decisions and sensitivity analysis
• Hardware integration
• Time and budget constraints
• Project management
• The goal is the same: maximize points by minimizing lap time

12.05.2019 Akademischer Motorsportverein Zürich | Angelo Corradi 33

driverless.amzracing.ch