TITLE

Contents:

• Motivation
• Past work
• Methodology
• Results
• Conclusions
• Implementation details
• References

Motivation

With the advancement in the field of Artificial Intelligence ,the research has been improved on tracking the vehicle in congestion road scenary traffic situation.With one step further,the behaviour of a car has been found in a petrol pump model by image processing sequence . The behaviour which hsa been found , has to be checked by executing the behaviour by sending back to the system.My project will be the first step for the above said problem.i.e Given the behaviour or the desired behaviour chosen by an agent himself wrt other agent ,the car will have to be simulated.The behaviour will be given as the position through which the vehicles could manouver themselves avoiding static and dynamic obstacles using potential fields as their guide.

Past work

• The paper summitted by Dr. Hans helmut nagel inwhich the behaviour of car has been found in a petrol pump model by image processing sequence.
• significant contributions have been made by many researchers like Dr. Latombe on Obstacle avoidance and Path planning. Also an M.Tech thesis was submitted by Mr. Shami Gupta (IIT Kanpur), which considered the movement of two cars and the course project for AI course was submitted by Groover's team (IIT Kanpur). In both the cases , the car is controlled by the increment position in x & y direction which is the funtion of potential field.
• Finally an M.tech thesis going to be submitted very soon by Mr.suman ojha(IIT Kanpur). in which the car movement is controlled by the accl of the car ,wherein the velocity is the function of the potn field. For other references related to this field, refer the Bibliography at the end of the report.

Methodology

Input :

The map of a petrol pump shown in fig is made by trapezoidal Segments and Junctions which joins the Segments. This representation is powerful enough to approximate real maps. the notion of a nominal path for vehicle is also given to guide it. The vehicle constitutes of a Driver (described in detail below) which alters the vehicle's runtime state with the objective of achieving the nominal path.

Driver :

Degrees of freedom of a car include:

• Front axle velocity
• Steering angle

The driver is posed with the problem of altering the state of the car inorder to achieve the nominal path and avoid the obstacles (static \& dynamic). Each degree of freedom can be altered in three different ways (increased, decreased or no change), thus reuslting in 9 possible future configurations. A potential is associated with each configuration of the car in such a manner that if it is minimized then the car shall try to achieve the desired path. Thus the job of the driver is to project the car in each of its 9 possible future configurations and take a decision by minimizing the potentials calculated. The increment (x,y) coordinates of the car are a result of the decisions the driver has made from the time the car started to move.

Calculation of the Potential

The potential felt by a vehicle is affected by :

• Road sides

The road sides are associated with the car by means of the segment boundaries associated with the car's current segment. These play a part only when the car is very close to them (Here the function x to the pow 4 is used to ensure this)

• Visible obstacles in its line of motion:

If the obstacle is in the same segment, ahead of it and is in the line of sight of the vehicle, it is visible to the vehicle.

Simulation of car2:

1. The next position will be given as the next behaviour to be executed after the car2 executing the previous behaviour which means reaching the previous position
2. If the agent of car2 wants to execute the behaviour autonomously (i.e without giving the position ) ,the agent will select the next position autonomously. This is done by selecting the middle pt of the following segments as the subsequent position.
3. If both the cars enters the pump at the same time from the side which is oppssite to one another ,car2 will enter left or right lane based on the behaviour of the car1 inorder to avoid collision.
4. If the car1 enters before car2 and stops at right ar left lane then car2 will be selecting the left or right lane wrt car2 on lanes..

Results

I have been made a robust program at my level best using the theory described before. My implementation is programmed in Borland C++. The simulation works reasonably well for many different cases mentioned in the above section. The movement of the vehicle is quite smooth. To achieve this smoothness an attractor runs on the nominal path at some distance away from the vehicle and attracts it, thus resulting in its asymptotic movement towards the nominal path.

Conclusions

• [L]The car is controlled by increment position in x,y direction

[F] This can be done by controlling the acceleration of the car as mentioned in the past work sction.

• [L]I have implementd only limted case with not more than two cars

[F] With more than two cars the more general cases such as if two cars enter into diff lane from tord from sides opposite to one another ,then the third car should follow the path of the car whose direction is same as third car

• [L]The time between the starting of car from current position to next position

Implementation details

• create a file named "pump_map.dat"which describes the mapping of petrol pump contains the coordinates of the various segments.
• create a file named " car.dat" & "car1.dat" and "car2.dat" describes the starting position and dimension of the car,various data required by the car and nompath throug which the car will be moved.
• finally a file called "pot_par " contains the various potential parameters required for the potential field method

Reference

1. Mr. Shami Gupta and Dr. Amitabha Mukerjee, : An M. Tech thesis Microscopic Simulation of Congested Traffic Flow using Potential Field Models. Dept of Mech Engg and Center for Robotics, Indian Institute of Technology, Kanpur, INDIA . 2.Groover,anoop,sachin and Dr.Amitabha Mukerjee:Course project Simulation of Congested Traffic Flow using Potential Field Models. Dept of Mech Engg and Center for Robotics, Indian Institute of Technology, Kanpur, INDIA
3. Barraquand, J. and Latombe, J-C., 1991. Robot motion planning : A distributed repre- sentation approach. International Journal of Robotics Research, Vol. 10,No.6, pp 628-649. 1991.
4. Fraichard, Th. and Laugier, C., 1991. On line reactive planning for a nonholonomic mobile in a dynamic world. Proceedings of the IEEE International Conference on Robotics and Automation, pp. 432-437. 1991.
5. Fraichard, Th., Scheuer, A., 1994. Car-Like Robots and Moving Obstacles. Proceedings of the IEEE International Conference on Robotics and Automation,May 8-13 1994, 6 Pages.
6.Huang, T., Koller, D., Malik, J., Ogasawara, G., Rao, B., Russell, S. and Weber, J., 1994 Automatic Symbolic Traffic Scene Analysis Using Belief Networks.Conference Proceedings of AAAI, pp. 966-972. 1994.
7.Hwang, Yong and Ahuja, N., 1992. A potential field approach to path planning. IEEE Transactions on Robotics and Automation, Vol. 8,No.1, pp. 23-32. 1992.