Skip to Content

Partial Autonomous Vision-Based Fuzzy Control Guided Car

Public transports take the same path every hour and every day. The huge amount of traffic inside cities nowadays and cars parked everywhere in both sides of the streets make it almost impossible to automate these kinds of vehicles using the traditional approach of lane detection. Furthermore, the streets are full of shadows produced by trees, buildings, and other city structures causing large brightness changes in the images. These kinds of visual effects are usually detected like occlusions by visual algorithms. This work presents a low cost approach to a visual-based partial autonomous vehicle robust against variation in brightness. The presented system just has one camera and doesn’t use any GPS information. A visual system has been developed to follow a predefined path and to get information about the vehicle position. The developed control system has been tested in different city situations with excellent results.

 

The camera is inside the metal structure locted in front of the vehicle. The field of view is limited to the size of this structure being 50 x 30 cm. This measure is enough to the system can drive itself reaching 50kph and going inside a circuit with city like curves of 11 and 20 meters of radio. A line is painted on the street and is detected by a visual algorithm. A binary type of mark was design to give extra information to the system. These marks give location information to the vehicle so the system can know if it close to a curve, what type of curve it is. This information is added to the control system by an offset. The control system is composed by a PD like Fuzzy controller implemented usin the Miguel Olivares' Fuzzy Software (MOFS). This controller has 49 rules that were set by heuristic methods and with a learning algorithm which use the information acquired by human driver performances. This controller has an extra integral block designed using classic control methods.

In the Figure below is shown the circuit used to test the intelligent system. It has an oval form with one curve of 11 meters of radius and another one with 20 meters of radius.

Fig. 1. Sat-view of the circuit used to test the system at the INSIA (Madrid-Spain)

 

 

 

 The system was tested for long time test to evaluate the correct behavior of the system. Here is presented one of this test in which the intelligent system was working during 23 minutes covering 30 laps of the circuit. The speed of the vehicle was set to 15 kph and cover almost 6 km. The Root Mean Squared Error (RMSE) was used to evaluated the test. The final RMSE of this test was 4.015 cm.

Figure 2 shows the measure of the speed during the test. The speed was set constant for this test.

Fig 2. Vehicle Speed in the 30 laps test.

Fig. 3. Movements of the steering wheel during the 30 laps.

Fig. 4. Error during the 30 laps with a full RMSE of 4.015 cm.