Towards an Intelligent Vehicle Model

[SRGRDGZ]

Contact Information:

Country: France
Year Submitted: 2018
University: Paris-Saclay University - University Institute of Technology of Orsay
List of Team Members (with year of graduation):

Sixtine Peniguel, Willieam Verlinde and Quentin Duparc

Students of the Applied physics and metrology Department, Sept 2019

Faculty Advisers: Sergio Rodriguez and Bastien Vincke
Main Contact Email Address: sergio.rodriguez@u-psud.fr / firstname.lastname@u-psud.fr

---

Project Information:

Title: Towards an Intelligent Vehicle Model
Description: This project is intended to design and to implement a demonstration pilot of an intelligent vehicle model. The proposed system architecture is composed of a distributed perception system tightly coupled to a centralized decision module.
Products:

• LabVIEW :
  • Robotics and Vision Toolkits
  • Development environment
• NI myRIO
• NI myRIO mecatronics kits
• Other hardware: ultrasound range finders, mbed micro-controller Nucleo F303K8, 

The Challenge:

In the last 5 years, an increasing deployment of intelligent transportation systems (ITS) have been observed in our society. Such transportation technology is not only aimed at reducing the important number of road accidents (e.g. active safety and driving assistance) but also at improving the comfort of road users (e.g automatic driving).

Our interest and efforts are focused on the design and the implementation of the embedded sensing system of an intelligent vehicle. To this end, we propose to implement an intelligent vehicle concept model (1:10 scale, see Fig. 1) that integrates similar perception functions to those of an autonomous car. Our reported end-to-end perception function implementation involves sensor technology choice, sensor error modeling, data acquisition design, data transfer and results displaying. The addressed functions are obstacle detection, line following and trajectory estimation.

Figure 1. Intelligent vehicle concept model

Project experiences will help us to better understand the perception capabilities and requirements of such a complex system as well as its energy management mechanisms and its onboard communication strategies.

The Solution:

Our solution is implemented on a 1:10 electric radio-controlled model car (see Fig. 2 and Fig. 3). The model integrates a servo-motor that allows us to control the steering angle of the vehicle. The longitudinal motion of the car is controlled by means of a geared DC motor coupled to a 4x4 traction mechanism.

Our contributed solution is structured following an electronic distributed architecture composed of 4 embedded modules also called Electronic Control Units (ECU) and an onboard main processing unit (i.e. myRIO). This architecture communicates through a CAN bus link (100 kbits).

Figure 2. Vehicle components overview

Figure 3. Intelligent vehicle model (Left-view)

Hereafter, an overall functionality description of each ECU is provided.

• Line following

This ECU is composed of an MBED micro-controller and a commercial line detector sensor integrated into an own-developed electronic card. Line detector is composed of 5 infrared detectors disposed in a row. Their numerical output indicates the presence/absence of a reflective surface (i.e. line).

The micro-controller performs a periodic acquisition of sensors measurements. Based on line detector measurements, an algorithm infers the correct steering angle for lane keeping. Then, pulse modulated signals for steering and electric throttle are applied.

The width of the detected line is employed to control longitudinal speed (i.e electric throttle). That is, larger the line is, slower is the speed of the vehicle.

• Obstacle detection

For obstacle detection, two sensing technologies were studied, ultrasound and infrared range finders. Both technologies provide different advantages and drawbacks. The ultrasound range detectors were selected since they cover a bigger field of view. In order to detect obstacles in front of and behind the vehicle , two sets of 3 ultrasound detectors were installed in the car's bumpers.

Since ultrasound range is as a time-of-flight measure, their operation requires a sequential acquisition to avoid a mutual interference between sensors. This constraint is integrated on the micro-controller as well as the sensors calibration. Obstacle detection results are finally represented into a discrete map of the vehicle surroundings (see Fig. 4) that is transferred as a CAN bus message to the main processing unit.

In presence of an obstacle close to the vehicle, an emergency braking message is emitted so as to set the vehicle speed to zero.

Figure 4. Obstacle detection example

• Trajectory estimation

The vehicle's trajectory is estimated by performing dead-reckoning. To this end, a quadrature encoder of the geared DC motor is used as a measure of the car speed. It is worth noting that it is necessary to take into account the influence on the vehicle speed of the 4x4 traction mechanism and the gear reduction.

A gyroscope located in the middle of the rear traction axle is employed to infer the relative vehicle orientation.

A micro-controller is dedicated to perform simultaneous linear speed and yaw rate measurements. They are integrated over time using the Ackerman steering model. As a result, the relative position of the vehicle with respect to its starting point is reconstructed.

• Main Processing Unit

The main processing unit onboard the model is a myRIO module. This unit access to the CAN bus and retrieves all data provided by ECUs. High-level and HMI functions are implemented on this unit. Thanks to the WiFi capabilities of myRIO, the state of the vehicle can be monitored in real-time.

Demonstration video