Vehicle through Expressway Toll Gate Application
In order to validate the proposed method, we recreated by simulation the toll gate inNagoya expressway, as shown in Fig. 7. In the simulation, all the vehicles (including the host vehicle) choose randompaths from the combination of the nodes associated to four start lanes, six entrance toll gates, and four exit destinationlanes, represented by the orange dots in Fig. 7. The total of 96 possible paths accounts for the potential complex trafficsituations of surrounding vehicles at the toll gate area. In every simulated iteration, we randomized the initial speed ofall vehicles from 0 km/h to 20 km/h, with random acceleration and deceleration between 1m/s2to -2.5 m/s2,and wesetthe velocities of all vehicles between 0 km/h and 60 km/h at any time. All the vehicles drive under the restriction ofthe vehicle kinematic model and only the host vehicle is controlled by the proposed method
In this simulation, various driving scenarios are generated from randomized conditionsto validate the performance and robustness of the proposed decision tree based method. We ran our MATLABcoded program for more than 100 times, and successfully reached the destination onevery occasion, showingthe robustness of the method. As a showcase, the following three scenarios are proposed, focusing on the toll gate exit area. Toll Gate Application For Mobile Based Payment In these three scenarios, the surrounding vehicles path, ini-tialspeed and maneuver (acceleration/deceleration) are se-lectedrandomlyto simulate different driving patterns at toll gate. The proposedmethod successfully generates different action plans for each scenario,dependingon the surrounding vehicle maneuvers. Figure8 shows the results of the decision tree for maneu-veringthe host vehiclein Scenario A. As shown in Fig. 8, the host vehicle selects thedecelerationtoavoid collision as thehost vehicle is unable topass the nearest near collision point by acceleration. In scenario A, the generated action plan is “Decelerate at First Near Collision Point”.
The path of the host vehicle connecting the start and end position is shown by the red linein Fig. 9(a). At every time step, the host vehicle predicts near collision points on its path according to the speed andheading of the surrounding vehi-cles. If the surrounding vehiclescross the path withinthe maximum allowed time tocollision𝑡𝑐(in this simulation the 𝑡𝑐= 5 seconds),the predicted near collision point will be added to a queue in an order according to its distance to the host vehicle. https://codeshoppy.com/ In Fig. 9(a), we use rectangles to represent ve-hicles and arrows to representtheir trajectories. The red rectangle represents the host vehicle, and the other colors repre-sent the surrounding vehicles.As shown, the host vehicle successfully predicts the near collision points (shown by the orange triangle)according to the speed and position of the surrounding vehicles at every time step. According to the results of theproposed decision tree, the host vehicle reduces its speed to give way to the yellow vehicle that is merging into its path.
The path of the host vehicle is shown in Fig.11(a). In the first time step, the host vehicledetects a near collision point, which can be passed without collision through acceleration. After passing the first nearcollision point, the host vehicle decelerates since the second collision point cannot be avoided through acceleration.
Scenario C is the case where the host vehicle finds that all the near collision points can bepassed through acceleration from the start to the end. The decision tree is shown in Fig.12. The generated action plan is“Accelerate at first to the second near collision point.” Note that the third near collision point is the imaginary pointadded to complete the algorithm, but the deceleration at the third near collision point will not be per-formed.The simulation results are shown in Fig.13. The path of the host vehicle is shown in Fig.13(a). The host vehicle detects near collisionpoints along the path, but recognizes that it can pass allof them safely. Figure 13(b) shows the speed and accelerationprofiles of the host vehicle.Since all near collision points can be passed, the host vehicle drives toward its destinationwith the maximum preset acceleration till it reaches a speed of 60km/h, which is the limited speed on the ramp
In this work, we have proposed a decision tree based method for controlling the speed of the host vehicle in ordertosafely drive across the wide and unstructured toll gate region with complex traffic situation. Our proposed method has low calculation cost and more human-like behavior compared to other classical path planning methods,due to the predefined path and limited extension of the decision tree. Our decision tree based method allows the host vehicle to accelerate to pass several vehicles at once, resulting in minimum changes in the acceleration or deceleration to maintain the comfort of pas-sengers with safely reaching the destination in the fastest manner. We validated robustness of our method through var-ious simulation scenarios recreatingthe complex traffic situa-tion of the toll gate. The results showed that the proposed method correctly prevents the hostvehicle colliding with surrounding vehicles through deceleration or acceleration, in all simulated scenarios.For the future work, we considerimproving the limitation of the current algorithm byimple-menting path correction algorithm for those cases in which a static object is detected on the predefined path. Inorder to further consider the application of this algorithm, we also consider applying our method to similar complexscenarios, such as a multilane intersection or a parking area with heavy traffic. Read more
codeshoppy 